ComfyUI Auto Installer: 1-Click Windows, Linux & Cloud With Presets (Patreon)
Downloads
Content
ComfyUI auto installer that sets up Sage Attention, Triton, xFormers and Flash Attention in one click, with ready presets for image, video and lip sync models.
Installs Torch 2.13 and CUDA 13 with FaceID, IP-Adapter, InsightFace, Reactor, DeepSpeed, MSLK and TorchAO, including RTX 5000 Series support. Automatic installers for Windows, RunPod, Massed Compute, SimplePod and Linux; the V154 update adds a 3-step LongCat-Video-Avatar 1.6 talking video preset.
Patreon exclusive posts index to find our scripts easily, Patreon scripts updates history to see which updates arrived to which scripts and amazing Patreon special generative scripts list that you can use in any of your task.
Join discord to get help, chat, discuss and also tell me your discord username to get your special rank : SECourses Discord
Please also Star, Watch and Fork our Stable Diffusion & Generative AI GitHub repository and join our Reddit subreddit and follow me on LinkedIn (my real profile)
=======
Latest zip file (Torch 2.13+CUDA 13) : ComfyUI_V154.zip
Tutorials
ComfyUI Master installation and how to use tutorial (17 October 2026) : https://youtu.be/nxsCBTcCVp4
How to update ComfyUI mini tutorial (31 October 2026) : https://www.youtube.com/watch?v=sdD3--WbHZk
SwarmUI Complete Master Tutorial with ComfyUI Backend (28 October 2026) : https://youtu.be/0HmIeeI1iBo
10 January 2026 SimplePod & Windows Tutorial : https://youtu.be/yOj9PYq3XYM
10 July 2025 RunPod + Massed Compute tutorial : https://youtu.be/8cMIwS9qo4M
Massive Models Downloader (use this to download core bundles to not have any issues with presets): https://www.patreon.com/posts/114517862
Bat files auto starts with --enable-triton-backend and --use-sage-attention if you don't want modify bat files (Windows_Run_GPU.bat)
Add this to your SwarmUI backend as well : --enable-triton-backend
If your GPU starts using shared VRAM for any reason, add this command to make it avoid that --reserve-vram 3
This command will preserve 3 GB VRAM for other tasks
It is like --use-sage-attention command
If you are on high VRAM GPU
--gpu-only : keeps everything on GPU
--highvram : keeps models on GPU
--cache-none is useful when you work with dual models or keep switching between models
--novram is useful when you are getting OOM, especially longer video generation like LTX 2, it reduces VRAM usage and can be used together with --cache-none
Some VRAM savings --disable-smart-memory, --lowvram
If you are getting stuck or OOM use this: --disable-smart-memory and --disable-comfy-compiler
We also have published amazing video and images comparison slider app : https://www.patreon.com/posts/133935178
Use Massed Compute installer for local Ubuntu / Linux machine
Now it will auto setup everything
Windows Requirements
Please use latest Windows Installer tutorial and zip file literally 1-click Bat file
1-time only necessary : https://youtu.be/ptuLL0C72Ms
Setup zip file : https://www.patreon.com/SECourses/posts/1-click-windows-172290854
18 November 2026 Update V154
Tutorial video : https://youtu.be/b_3BnPXouGc
New: LongCat-Video-Avatar 1.6 Really Fast. A new preset turns one photo and any audio into a lip-synced 720p talking or singing video of any length, with 3 of the official 8 distilled steps. And one new premerged model now runs every LongCat preset.
The complete Presets folder of V154, the new preset marked NEW and the updated one UPDATED:
Every example in this post is a real LongCat-Video-Avatar 1.6 result, nothing retouched.
NEW PRESET — LONGCAT AVATAR 1.6, 720P, 3 STEPS REALLY FAST
"LongCat Avatar 1.6 - Image + Audio - 720p 3 Steps Really Fast - Premerged DMD + OPSD - 261118"
Your photo and your voice become a 720p talking or singing video of any length, with 3 of the official 8 steps. The last segment is made only as long as your audio needs, and the faster INT8 attention is already on. All models come in the LongCat-Video-Avatar 1.6 Core Bundle, first in our downloader:
The bundle's new default model is our premerged INT8 ConvRot: both distillation LoRAs are merged in, so no LoRA work happens while sampling, and one model runs every LongCat preset. The plain INT8 ConvRot and BF16 models stay available in the downloader.
One photo and a 15-second voice recording, made with the new preset at its default 720p:
The same man, glasses, T-shirt and room from the first second to the last, while his face, mouth and hands move with the words.
The same moments of the official 8 steps and the new 3-step preset, made from the same photo, speech, seed and size:
In our test with this 15-second 720p clip, the 3-step preset made the video about 2.9 times faster than the official 8 steps, with the same person, glasses, room and lip sync.
All our extras are kept: the prompt box with + Prompt and folder batch, soundtrack load and trim, the optional photo, the optional LoRA stack, live preview, the mouth enhancement switch, and a note beside every setting:
Upload your audio and photo; everything else is already set.
Where to find it: open the preset, upload your audio and photo, and press Run:
The finished video plays in the preview with your audio and is saved in ComfyUI's output folder under video/LongCat_Avatar.
UPDATED PRESET — LONGCAT AVATAR 1.6, 8 STEPS
"LongCat Avatar 1.6 - Audio + Image or Text To Long Lip Synch Video - 8 Steps - 261117"
The official 8-step recipe now runs on the same premerged model. Same look: on the same photo, audio and seed it makes the same video as before, and you keep only one LongCat model on disk.
HOW TO UPDATE
Download the latest ComfyUI_V154.zip from this post and extract it over your installer folder.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat.
Download the LongCat-Video-Avatar 1.6 Core Bundle with our SwarmUI Model Downloader V195.
Restart ComfyUI and refresh the browser tab.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
17 November 2026 Update V153
Tutorial video : https://youtu.be/b_3BnPXouGc
New: LongCat-Video-Avatar 1.6. Turn one photo and any audio into a lip-synced talking or singing video of any length, on your own GPU. The video follows your whole audio: it is made in segments of about 4 seconds that continue each other, with your photo as the identity reference, so a one-minute talk stays the same person in the same room from the first to the last frame.
No photo? Then the scene comes from your audio and one sentence of text.
The complete Presets folder of V153, the new preset marked NEW:
Every example in this post is a real LongCat-Video-Avatar 1.6 result, nothing retouched.
NEW PRESET — LONGCAT-VIDEO-AVATAR 1.6, LIP SYNC VIDEO OF ANY LENGTH
"LongCat Avatar 1.6 - Audio + Image or Text To Long Lip Synch Video - 8 Steps - 261117"
Your photo and your voice become a talking or singing video, as long as your audio. Upload the audio, add a photo and one sentence of scene if you like, and press Run. All models come in the new LongCat-Video-Avatar 1.6 Core Bundle, the first bundle in our downloader:
The default model is our INT8 ConvRot version: about 17 GB instead of almost 30 GB for BF16, as the downloader shows them, and its sampling runs 1.7x faster than BF16 in our RTX 5090 test (one 93-frame segment at 512 x 768, 8 steps). BF16 stays available in the downloader.
The same presenter photo and the same 10-second narration at 1280 x 1920, made with LongCat-Video-Avatar 1.6, our MiniMax H3 lip sync preset and AvatarForever:
Same photo, same audio, same size: only the model changes. All three keep the man, his glasses and the library of the photo, and LongCat-Video-Avatar 1.6 also moves his hands as he talks.
Mouth Enhancement is one switch: our CodeFormer mouth restoration runs on every frame, for cleaner, more even teeth and smoother lip edges. Turn on save original too and you also keep the untouched video next to it. The same frames zoomed 4x, off on the left and on on the right:
Vocal separation is on by default, as in the official pipeline: background music does not move the mouth, and the saved video keeps your original audio.
Real jobs at 720p, each from one photo and one voice line: a chef presenting the special of the day, a real-estate agent showing a listing, a family doctor with a health tip and a fitness coach:
A 58-second talk made in one run. Fourteen frames from the first second to the last: the same man in the same room the whole time, while his face and hands keep moving with the words:
The same one-minute talk made with AvatarForever too: the same photo, the same narration, the same seed and the same 512 x 768 size, with mouth enhancement on in both. The same seven moments of both videos:
Both keep the man, his glasses, his T-shirt and the room for the whole minute. LongCat-Video-Avatar 1.6 moves his head and hands more as he talks, and AvatarForever keeps a steadier framing. The full side-by-side video with sound is attached to this post.
Long videos also get faster with Condition Cache, an exact speed-up that switches itself off at very large sizes. ComfyUI's low-VRAM offload works with our nodes.
Our extras are all in the preset: + Prompt makes one video per prompt, a folder batch pairs name.txt with name.png and name.wav by itself, a live preview shows the video while it renders, optional LoRAs are one switch away, and every setting has a note beside it:
Pick speech or singing in Performance; everything else is already set to the official 1.6 recipe.
Where to find it: open the preset, upload your audio and photo, write one sentence of scene and press Run:
The finished video plays in the preview with your audio and is saved in ComfyUI's output folder under video/LongCat_Avatar.
HOW TO UPDATE
Download the latest ComfyUI_V153.zip from this post and extract it over your installer folder.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat. It installs and updates every node our presets use.
Download the LongCat-Video-Avatar 1.6 Core Bundle with our SwarmUI Model Downloader V194.
Restart ComfyUI and refresh the browser tab.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
11 November 2026 Update V151
Installers made better, more robust and more simplified
Now we have only single bat file to install and update everything you need : Windows_Install_Or_Update_ComfyUI.bat
For RunPod and SimplePod just follow:
RunPod_SimplePod_ComfyUI_Instructions_READ.txt
For Massed Compute and local linux PCs just follow:
Massed_Compute_Instructions_READ.txt
14 November 2026 Update V152
New: MiniMax H3 Image with KJNodes' empty-image method, for cleaner skin, hands and fine detail. A new Text To Image preset shows the model an empty picture during the first steps, a method from kijai's KJNodes. On many prompts it gives more natural skin, cleaner hands and sharper small objects than the standard preset, at the same 16 MP and with the same prompt enhancer.
The prompt enhancer is now repeatable, too: a fixed seed gives the same caption and the same image every time, in every MiniMax H3 Image preset.
The complete Presets folder of V152, the new preset marked NEW and the improved ones UPDATED:
Every image in this post comes from a real run of the shipped presets with their default settings.
NEW PRESET — TEXT TO IMAGE WITH KJNODES EMPTY IMAGE
"MiniMax H3 Image - Text To Image 16 MP - KJNodes Empty Image - 6 Steps - 261114"
Cleaner skin, hands and fine detail from the same short prompt. Write your idea and press Run, exactly like the standard Text To Image preset. It uses the same models, all in the MiniMax-H3 Core Bundle:
Already using MiniMax H3 Image? There is nothing new to download. The MiniMax-H3 Low VRAM Bundle has the smaller W4A8 versions.
The same caption and seed, the standard preset on the left and the new one on the right, with details at their real pixels below. A watchmaker at his bench: the watch case, the gears and the loupe come out clean:
A chef plating a dish for a restaurant ad, with clean hands and a detailed plate:
An olive farmer for a brand campaign: natural, weathered skin instead of a crackled texture:
The empty picture also changes the composition, like a new seed, so keep both presets and use the one you like for each prompt.
Where to find it: open the preset, write your idea in the gallery box and press Run:
The box under the result shows the exact prompt the model received. This preset makes new images; to edit your photos, use the Image Edit preset.
UPDATED — MINIMAX H3 IMAGE TEXT TO IMAGE AND IMAGE EDIT
A fixed seed repeats your image: the prompt enhancer now writes the same caption for the same seed, so the image repeats too.
Tidier layout: the model loaders no longer overlap the nodes next to them.
UPDATED — KREA 2, PHANTOM WAN, LTX AND FLUX PRESETS
Krea 2: all three presets use the BF16 text encoder of the Krea 2 Core Bundle under its own file name. Already using Krea 2? Download the Krea 2 Core Bundle once more with SwarmUI Model Downloader V193 to add it. Raw Base also has a negative prompt box now.
Phantom Wan 14B Character Reference and LTX2.3 Licon MSR V2: the reference image switches open with the right on and off settings.
FLUX 2 Low RAM: the VRAM cleanup before sampling is optional now and off by default; switch it on if you run out of VRAM.
LTX2.5 Text or Image To Video and FLUX 3 Klein Advanced Image Edit: they open without missing example images.
ONE INSTALLER FOR EVERY PRESET
Windows_Install_Or_Update_ComfyUI.bat installs every custom node our presets use, and so do the RunPod and Massed Compute installers: no bundle menu, no separate node installer.
IP-Adapter and FaceID models now come from the new IP-Adapter And FaceID Bundle in our SwarmUI Model Downloader.
Installer fixes: a moved, renamed or copied ComfyUI folder gets a working environment again, and a failed step is reported as an error.
HOW TO UPDATE
Download the latest ComfyUI_V152.zip from this post and extract it over your installer folder.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat. It installs and updates every node our presets use.
If you do not have it yet, download the MiniMax-H3 Core Bundle (or the Low VRAM Bundle) with our SwarmUI Model Downloader V190 or newer.
Restart ComfyUI and refresh the browser tab.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
10 November 2026 Update V150
Presets Folder Content In Latest Zip File
MiniMax H3 Image is here: 16-megapixel images in 6 steps. Two new presets turn a few words into images up to 6272 pixels wide and edit your photos with up to 16 images at once. MiniMax's own prompt enhancer runs locally inside the presets: a short idea in any language becomes the detailed caption the model was trained on, and the enhancer even picks the best shape for your image.
Real runs of the new presets:
Every image in this post comes from a real run of the shipped presets with their default settings.
The complete Presets folder of V150, the new ones marked NEW:
To run them, download the MiniMax-H3 Core Bundle. It now includes MiniMax H3 Image, its VAE and the prompt enhancer model; the text encoder is the one MiniMax H3 video already uses, so H3 users download only the new files:
Both new presets use this one bundle. The MiniMax-H3 Low VRAM Bundle has the smaller W4A8 versions.
NEW PRESET — TEXT TO IMAGE AT 16 MEGAPIXELS
"MiniMax H3 Image - Text To Image 16 MP - 6 Steps - 261110"
Type an idea, press Run, get a 16-megapixel image. The prompt enhancer writes the full description and chooses the aspect ratio, from tall posters to 21:9 panoramas. A snow leopard from one sentence, at 5024 x 3360:
It writes text you can read. A travel poster with two exact lines of lettering, 3072 x 5472:
Panoramas come out in one piece, with no upscaler. Asked for a wide view, the enhancer chose 21:9, 6272 x 2688; below it, three details at their real pixels:
Fine detail at 4096 x 4096. Open it full size and read the screw slots and the jewels:
Where to find it: open the preset, write your idea in the gallery box and press Run:
The box under the result shows the exact prompt the model received. Set the enhancer to Off to use your own long caption; our new MiniMax H3 Image LLM guide in the zip makes ChatGPT, Claude or Gemini write one for you.
1 MP VS 4 MP VS 16 MP
Why 16 MP is the default. The same caption and seed at three sizes, with the face from each cropped to the same share of the frame:
16 MP has sixteen times the pixels of 1 MP, and it uses them: fur, whiskers and frost that 1 MP can only blur. It stays fast, too. Seconds per image on an RTX 5090, also with ComfyUI limited to 24, 16 or 12 GB of its memory:
A 16 MP image takes about 16 seconds with all memory and about 20 seconds when ComfyUI may use only 12 GB; the rest streams from system RAM. Lower the size in 02 OUTPUT SIZE whenever you want a quick draft.
NEW PRESET — IMAGE EDIT WITH UP TO 16 IMAGES
"MiniMax H3 Image - Image Edit Up To 16 Images - 6 Steps - 261110"
Edit your photo with one sentence. Add the image, write the change in any language and press Run. The prompt enhancer looks at your images and writes the full edit description, and the result keeps your photo's shape at 16 MP. A moss-covered wreck comes back to life, while the forest, the branches and the camera stay:
Combine many images. An empty room and three product photos, named @image1 to @image4 in the instruction: the exact armchair, lamp and plant land in the unchanged room:
Where to find it: open the preset, add your images to the gallery, write the change and press Run:
Leave the gallery empty and it works as a text to image preset, so both presets do both jobs.
BUILT IN, LIKE ALL OUR PRESETS
Batches: + Prompt adds more prompts in a row, and a folder of .txt files runs one image per file.
The prompt enhancer in three modes: official with reasoning (best), fast without reasoning, or off.
Optional LoRA stack, shared seed and images per prompt.
Less memory: pick the W4A8 model in the same loader; BF16 is there for original precision.
HOW TO UPDATE
Download the latest ComfyUI_V150.zip from this post and extract it over your installer folder.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install bundle 100.
Download the MiniMax-H3 Core Bundle (or the Low VRAM Bundle) with our SwarmUI Model Downloader V190 or newer.
Restart ComfyUI and refresh the browser tab.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
4 November 2026 Update V149
Presets Folder Content In Latest Zip File
Edit real videos with MiniMax H3, and let the presets do the work. Six new presets replace the background of any video from one image, relight a whole clip, put a new face on someone, move a subject along a path you draw, upscale your video and replace a person or object while every other pixel stays as filmed. The Pika H3 Flex presets now write their own prompt when you leave it empty, and a long video can follow a story, scene by scene.
Real footage and real runs of the shipped presets:
Every result in this post comes from a real run of the shipped presets.
The complete Presets folder of V149, the new ones marked NEW and the updated ones UPDATED:
To run them, download the MiniMax-H3 Core Bundle. It includes the FaceSwap LoRA and the H3 latent upscaler; if you already have the bundle, only the missing files download:
All the new presets and the updated Pika H3 Flex presets use this one bundle.
BUILT IN — OUR LIVE TOKEN COUNTER
See what a run needs before you press Run. Our gallery counts the tokens of your text, references, audio and video as you work and shows them against the MiniMax H3 limit, with the size, the length and the mode beside them. It moves from green to amber to red as you get close, so you can shorten a clip or lower the size first. With long video windows it counts per window, because each window is one run:
Beside it in every preset: Load + trim to pick the part of your clip, folder batch with optional merging, and Continue From Last Video Frames.
NEW PRESET — BACKGROUND REPLACE: YOUR VIDEO, A NEW BACKGROUND, NOTHING ELSE
"MiniMax H3 Background Replace - Pika H3 Flex 6 Steps - 261104"
Fully automatic: give it your video and a background image, and press Run. The person is found by itself and stays exactly as filmed: every pixel of the face, hair, clothes and hands, the lip movements, the timing and the sound. Everything around them becomes the place in your image and moves with your camera.
A talking man on a grey wall goes to a tropical beach, and a walking man in New York goes to a futuristic city:
There is nothing to write or set. Add your two files, press Run:
Long clips work too; we tested 31 seconds (736 frames). The video is made in overlapping windows, so a long clip needs no more GPU memory than a short one, only more time. The result is about 1 megapixel; run it through the Video Upscale preset below for a sharper video. People of similar size all stay, and distant passers-by go.
NEW PRESET — VIDEO RELIGHT
"MiniMax H3 Video Relight - Pika H3 Flex 6 Steps - 261104"
Relight a whole video. Add your clip and pick one of 20 looks such as neon noir, golden sunset backlight or blue hour, or design up to three lights yourself with their direction, softness and colour. The people, the motion, the camera and the set stay.
The same street clip in neon noir:
The look, the lights and a preview sphere of where the key light comes from:
H3 follows the plan through the prompt, so it is not pixel-exact light control; try two or three seeds.
NEW PRESET — FACE SWAP
"MiniMax H3 Face Swap - Ref2VA 8 Steps - 261104"
Put a new face on the person in your video. Add your clip and a clear close-up of the new face. Clothes, setting and camera stay as in your clip, and the body motion follows it.
A selfie clip with a new red-bearded face:
Two inputs and Run:
H3 makes the whole clip anew, so the timing of the motion can drift a little. For your clip's exact timing and sound, use the Masked Edit preset below with target set to head.
NEW PRESET — PIKA H3 FLEX CUT AND DRAG MOTION
"Pika H3 Flex - Cut And Drag Motion To Video 6 Steps - 261104"
Draw a path on your photo and the subject follows it. Upload your photo and drag the path. The subject and the prompt are found and written for you. The subject is cut out and dragged along your path, and Pika H3 Flex turns that rough drag into a real video with moving light and sound.
From the photo, through the rough drag, to the finished drive:
Your photo, your path and Run:
Type a subject or a prompt whenever you want full control. Then the prompt model is not loaded at all.
NEW PRESET — VIDEO UPSCALE
"MiniMax H3 Video Upscale - Pika H3 Flex - 261104"
Fully automatic: give it your video and press Run. Your clip is encoded into MiniMax H3, enlarged by the H3 latent upscaler and refined by the last Pika H3 Flex steps, so the new detail is real video detail, not a sharpening filter. Your soundtrack stays exact.
Real footage reduced to 1280 x 704 and upscaled to 2560 x 1408, input on the left and result on the right, at the same moment:
Add your video and press Run:
What to expect: landscapes, animals and textures gain the most. H3 redraws fine detail, so small text and fine logos can change; on our harbor test a boat's small name came out garbled. One run makes the video at most twice as large.
NEW PRESET — GENJUTSU MASKED EDIT: CHANGE ONE THING, KEEP EVERYTHING ELSE
"MiniMax H3 Genjutsu - Masked Edit - Pika H3 Flex 6 Steps - 261104"
Replace a person or an object and keep the rest of your clip exactly as filmed. The street, the passers-by, the signs, the camera move and the sound stay untouched; only the target is regenerated and pasted back over your original frames.
A man on a New York street becomes a knight in armour, and the traffic behind him keeps moving:
Your clip and the new character are all it needs. The person is found and the prompt is written for you:
To replace something else, name it in target: a car, a dog, a man in a blue suit. For a face swap with your clip's exact timing and sound, set target to head and use a close-up of the new face.
UPDATED — PIKA H3 FLEX: LONG VIDEO THAT TELLS A STORY
"Pika H3 Flex - Text To Video 6 Steps - 261030"
A long video that develops instead of repeating. Switch on 9. LONG VIDEO, write your description, then a line with ---, then one scene, another ---, the next scene. Each part of the video follows its own scene.
Five scenes in one prompt: one continuous 20-second flight from a meadow to a sunset over the mountains:
The switch and the prompt:
About 4 seconds per scene works well. Without --- the preset works as before: every window reads the same prompt, so motion can repeat. Scenes work in text runs, not with reference media.
UPDATED — PROMPTS WRITE THEMSELVES
"Pika H3 Flex - First Last Frame To Video 6 Steps - 261030"
"Pika H3 Flex - References To Video 6 Steps - 261030"
Leave the prompt empty and it is written for you. A local Qwen3-VL 8B looks at your frames or reference images and writes the MiniMax H3 prompt. Type your own whenever you want full control, and that model is not loaded at all.
UPDATED — PIKA H3 FLEX: 2K
Sharper video with one switch. In the Pika H3 Flex presets, turn on 8. 2K UPSCALE and the video is refined at 2560 pixels with the same sound. Same seed, without and with the 2K pass:
Where to find it: in any of the three Pika H3 Flex presets, switch on 8. 2K UPSCALE, or 9. LONG VIDEO for clips over 15 seconds:
Both switches start off, so your runs stay as before until you turn them on.
HOW TO UPDATE
Download the latest ComfyUI_V149.zip from this post and extract it over your installer folder.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install bundle 100.
Download the MiniMax-H3 Core Bundle with our SwarmUI Model Downloader V189 or newer.
Restart ComfyUI and refresh the browser tab.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
30 October 2026 Update V147
Presets Folder Content In Latest Zip File
Pika H3 Flex: cinematic video with synchronized sound in 6 steps. Three new presets bring Pika Labs' new Pika H3 Flex to ComfyUI: text to video, first + last frame to video and references to video, each with dialogue, sound effects and music made together with the picture.
Our three examples: a ship in a bottle that sails out onto a stormy ocean, a blacksmith whose golden eagle lands on his anvil, and an airship captain made from just two images:
Every result in this post is a real run of the shipped presets on our RTX 5090.
The complete Presets folder of V147: 116 ready-to-use ComfyUI presets plus 3 API presets, the new ones marked NEW:
The presets cover Anima, AvatarForever, ERNIE Image, FastH3 Omni, Fibo 2.0, FLUX.1 Dev, Kontext and SRPO, FLUX 2, FLUX 3 Klein, Ideogram 4, Krea 2, LTX 2, LTX 2.3, LTX 2.5, MiniMax H3 with Genjutsu, MiniMax Music 3, MOSS-Image, Phantom Wan, Pika H3 Flex, Qwen Image, SDXL, Stable Video Infinity 2 Pro, Wan 2.1, Wan 2.2 and Z-Image.
NEW PRESETS — PIKA H3 FLEX: VIDEO WITH SOUND IN 6 STEPS
"Pika H3 Flex - Text To Video 6 Steps - 261030"
"Pika H3 Flex - First Last Frame To Video 6 Steps - 261030"
"Pika H3 Flex - References To Video 6 Steps - 261030"
Pika's tuned MiniMax H3, in 6 steps. A 5-second 1344 × 768 clip with stereo sound takes about a minute on an RTX 5090. Each preset switches by itself between the text and frames model and the references model.
To run them, download the MiniMax-H3 Core Bundle. It now includes the two Pika H3 Flex LoRAs; if you already have the bundle, only those 3.6 GB download:
First + Last Frame: give it two images and it makes the shot between them. Our ship in a bottle becomes the real ship on the open sea in one continuous camera move:
Where to find it: Workflows → SECourses → Pika H3 Flex - First Last Frame To Video 6 Steps - 261030. Upload your first frame in INIT IMAGE and your last frame in LAST FRAME, describe what happens and press Run:
The timing line Pika H3 Flex expects is added for you with the real length of your clip. Leave LAST FRAME empty for first frame to video.
Text To Video: one prompt gives the whole scene, with the blacksmith's line, the hammer, the hiss of steam and the eagle's wingbeats in the soundtrack:
References To Video: our captain image and sky harbour image went in, a walking, talking scene came out, with his beard, goggles, green scarf and compass kept:
Prompt batches, .txt folder batches, Init Audio, an optional LoRA stack and face refinement are built in. The MiniMax H3 prompt guide now covers Pika H3 Flex, so ChatGPT, Claude or Gemini can write your prompts.
HOW TO UPDATE
Download the latest ComfyUI_V147.zip from this post and extract it over your installer folder.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install bundle 100.
Download the MiniMax-H3 Core Bundle with our SwarmUI Model Downloader V187 or newer.
Restart ComfyUI and refresh the browser tab.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
28 October 2026 Update V146
Presets Folder Content In Latest Zip File
Anima character transfer, MOSS-Image layout control and stronger MiniMax H3 Genjutsu. Put one character into any scene with Anima, place and pose people with simple color and depth maps in MOSS-Image, decode FastH3 videos faster with LightVAE, and swap characters in your clips more convincingly with the new Genjutsu task adapters.
MOSS-Image Color + Depth Control turned the adapter author's two-person example maps into photos, and MOSS Image Edit then added snow and autumn without touching the two men:
Every result in this post is a real run of the shipped presets on our RTX 5090.
The complete Presets folder of V146: 113 ready-to-use ComfyUI presets plus 3 API presets, new ones marked NEW and updated ones UPDATED:
The presets cover Anima, AvatarForever, ERNIE Image, FastH3 Omni, Fibo 2.0, FLUX.1 Dev, Kontext and SRPO, FLUX 2, FLUX 3 Klein, Ideogram 4, Krea 2, LTX 2, LTX 2.3, LTX 2.5, MiniMax H3 with Genjutsu, MiniMax Music 3, MOSS-Image, Phantom Wan, Qwen Image, SDXL, Stable Video Infinity 2 Pro, Wan 2.1, Wan 2.2 and Z-Image.
NEW PRESETS — ANIMA: TAKE YOUR CHARACTER INTO A NEW SCENE
"Anima - Advanced Text To Image - 261028"
"Anima - Reference Character And Style - 261028"
One image of your character, any new scene. The Anima reference adapter carries the face, hair and outfit, and part of the drawing style, from your image, while your prompt picks the pose, place and light. Every final image is a clean 2048 × 2048.
To run them, download the Anima Core Bundle (3.84 GB) from the SwarmUI Bundles tab of our model downloader:
Our example: one character made with text to image, then four new scenes from that single image in one prompt batch:
Where to find it: Workflows → SECourses → Anima - Reference Character And Style - 261028. Add your character with + Add references, write the new scene, add more scenes with + Prompt and press Run:
Prompt batches, .txt folder batches and an optional LoRA stack are built in. The new Anima_Enchance_Prompt_Feed_For_LLMs.txt lets ChatGPT, Claude or Gemini write your prompts or a whole scene series.
NEW PRESETS — MOSS-IMAGE COLOR + DEPTH CONTROL: PLACE AND POSE PEOPLE WITH SIMPLE MAPS
"MOSS Image - Color + Depth Control - 261027"
"MOSS Image - Turbo 8 Steps Color + Depth Control - 261027"
Draw the layout, MOSS-Image follows it. A color map marks where each person or object goes, a depth map shows what is near and far, and your prompt says who they are. The Turbo preset makes a 1024 × 1024 image in about 7 seconds.
To run them, download the MOSS-Image Core Bundle. It now includes the color + depth adapter; if you already have the bundle, only that 0.29 GB file downloads:
The 50-step preset after the run that made the park photo at the top of this post:
Load your maps with choose file to upload in each map node. Part 8 of the MOSS-Image prompt guide writes the prompt for your maps.
UPDATED PRESETS — MOSS IMAGE EDIT: NOW ON COMFYUI'S NATIVE EDIT SUPPORT
"MOSS Image Edit - Advanced Image Edit - 261022"
"MOSS Image Edit - Turbo 8 Steps - 261022"
Both edit presets now run on ComfyUI's own MOSS-Image edit support and open with an empty gallery: add your photo, type the instruction, press Run. Up to 4 images, and the Turbo edit takes about 6 seconds. They use the same MOSS-Image Core Bundle.
NEW PRESET — FASTH3 OMNI REFERENCES TO VIDEO WITH THE LIGHTVAE DECODER
"FastH3 Omni - LightVAE References To Video 8 Steps - 261027"
The same FastH3 Omni video, decoded faster. LightVAE, a lighter video decoder from LightX2V, decoded and saved our 5-second video in under 6 seconds instead of over 8, with nearly the same picture.
To run it, download the FastH3 Omni Core Bundle. It now includes the LightVAE decoder and the Genjutsu adapters:
The preset after a real run:
Switch the decoder node to Original video VAE and press Run to compare: ComfyUI only decodes again.
UPDATED PRESETS — MINIMAX H3 GENJUTSU: CHARACTER SWAP AND MOTION TRANSFER ADAPTERS
"MiniMax H3 Genjutsu - FastH3 Omni 8 Steps - 261024"
"MiniMax H3 Genjutsu - Ref2VA 20 Steps - 261024"
Swap the person in any clip for your character, now with adapters trained for exactly this job. Genjutsu keeps the clip's motion, timing, camera and sound; the new Character Swap and Motion Transfer adapters by Alisson Anjos make the transfer much stronger. Clips with mono sound now work too.
The FastH3 Omni 8 Steps preset uses the FastH3 Omni Core Bundle above. The Ref2VA 20 Steps preset uses the MiniMax-H3 Core Bundle, which now includes the three Genjutsu adapters:
Our armoured warrior in a shadowboxing clip, made by both updated presets:
The FastH3 Omni Genjutsu preset after that run, with the new parts numbered:
Pick video background to keep the clip's scene or image to keep your character's background. Low VRAM mode is on by default with the same output.
HOW TO UPDATE
Download the latest ComfyUI_V146.zip from this post and extract it over your installer folder.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install bundle 100.
Download the core bundles shown above with our SwarmUI Model Downloader V186 or newer: Anima, MOSS-Image, FastH3 Omni or MiniMax-H3.
Restart ComfyUI and refresh the browser tab.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
24 October 2026 Update V145
Presets Folder Content In Latest Zip File
This is a MiniMax H3 Genjutsu update. V145 adds two Genjutsu presets that replace the person in any video with your own character: add your clip and one image of the new character, press Run, and the result keeps the clip's motion, timing, camera, length, aspect ratio and original sound. One preset uses FastH3 Omni at 8 steps, the other MiniMax H3 Ref2VA at 20 steps. V145 also adds a Genjutsu section to the MiniMax H3 prompt guide and brings the new Genjutsu nodes of our FoleyExtension.
Our test: this armoured warrior image and a 9.8-second phone clip of a man shadowboxing in a gym. Both new presets put the warrior into the clip, punch for punch and step for step, in the clip's portrait format and with its own sound:
Every result in this post comes from a real run of the shipped presets in ComfyUI started with our Windows_Run_ComfyUI_Default.bat: each preset was opened from the Workflows sidebar and run with the Run button, with its shipped settings and the example it opens with. The times are ComfyUI's times on our RTX 5090 for a first run, model loading included. We ran both presets again on the newest ComfyUI that the installer installs today and got the same videos, frame for frame.
Exact V144 → V145 zip audit:
2 new files: the two MiniMax H3 Genjutsu presets.
1 changed file: the MiniMax H3 prompt guide, with the new Genjutsu section.
No files were removed. The other 238 files are identical. The Genjutsu nodes arrive through Windows_Install_Or_Update_ComfyUI.bat, which installs our updated FoleyExtension.
Here is the complete Presets folder of V145: 108 ready-to-use ComfyUI presets, plus 3 API presets in the api_paid folder. The two new presets are marked NEW:
The presets cover image, video, lip sync, audio and music models: AvatarForever, ERNIE Image, FastH3 Omni, Fibo 2.0, FLUX.1 Dev, Kontext and SRPO, FLUX 2, FLUX 3 Klein, Ideogram 4, Krea 2, LTX 2, LTX 2.3, LTX 2.5, MiniMax H3 (now with Genjutsu), MiniMax Music 3, MOSS-Image and MOSS Image Edit, Phantom Wan, Qwen Image, SDXL, Stable Video Infinity 2 Pro, Wan 2.1, Wan 2.2 and Z-Image.
NEW PRESETS — MINIMAX H3 GENJUTSU: PUT YOUR CHARACTER INTO ANY VIDEO
"MiniMax H3 Genjutsu - FastH3 Omni 8 Steps - 261024"
"MiniMax H3 Genjutsu - Ref2VA 20 Steps - 261024"
How it works: SAM 3.1 finds the person in every frame of your clip, and only that person is colour-inverted. MiniMax H3 then copies the clip's motion, timing, camera and lip movements, but not the performer's face, hair or clothes. A local Qwen3-VL 8B (Auto Describe) looks at your character image and six frames of the clip and writes the prompt for you: the new outfit, the action, the setting, the lighting and the framing. The clip's soundtrack is locked into MiniMax H3's audio stream for lip sync, and the saved video carries the original sound.
Your inputs: the clip as @video1 and one image of the new character as @image1, added with Add references or Load + trim (to pick the part of the clip you want; 4 to 10 seconds works best, up to 15). The presets open with our example loaded; if those example files are not on your PC, remove the two cards and add your own.
The output follows the clip: the same length, the same aspect ratio at about 1 megapixel, and the original sound. Our 9.8-second 1080x1920 clip gave 226 frames (9.4 seconds) at 768x1344.
Prompt: keep the short default note, type your own notes (for example "He wears his red scarf."), empty it, or paste a complete MiniMax H3 prompt from ChatGPT, Claude or Gemini; a complete prompt is used as written.
Background: video keeps the clip's location, lighting and framing and replaces only the performer (character replacement); image keeps your character image's background and takes only the motion (motion transfer).
Target: who to replace, in plain words: person, or man in red:1 to pick one of several people.
Speed: FastH3 Omni 8 steps made the 9.4-second example in 4.4 minutes and Ref2VA 20 steps in 8.3 to 9.4 minutes, first runs with model loading on an RTX 5090.
Here is the FastH3 Omni preset right after the run of our example, with every part numbered:
After a run, CHECK THE MASK shows the performer in red in every frame, MOTION GUIDE shows what MiniMax H3 copies the motion from, and FINAL PROMPT SENT TO H3 shows the prompt Auto Describe wrote. The token meter under the prompt now follows Genjutsu too: it shows the clip's length and aspect ratio before you press Run.
UPDATED PROMPT GUIDE — MINIMAX H3 GENJUTSU
The MiniMax H3 guide in Automatic_Prompt_Generate_Enhance_To_Video_And_Image_Models_LLM_Instructions has a new Genjutsu section. Give it to an LLM that can watch video (for example Gemini) together with your clip and your character image, and paste its answer into the prompt box. It can write the exact spoken words for word-by-word lip sync.
HOW TO UPDATE
Download the latest ComfyUI_V145.zip from this post.
Extract it over your existing installer folder and overwrite the older files.
Close ComfyUI, and SwarmUI too if it uses this ComfyUI as its backend, then run Windows_Install_Or_Update_ComfyUI.bat. It updates ComfyUI and installs our FoleyExtension with the Genjutsu nodes.
If you have not done it since V144, run Windows_Custom_Nodes_Bundles_Installer.bat and install/update bundle 100.
Restart ComfyUI and refresh the browser tab.
Use our latest SwarmUI Model Downloader (SwarmUI_Model_Downloader_v185.zip or newer) and download the bundle from the SwarmUI Bundles tab:
FastH3 Omni Core Bundle (60.12 GB, 11 files) for the FastH3 Omni Genjutsu preset, or the MiniMax-H3 Core Bundle (79.91 GB, 19 files) for the Ref2VA 20 steps preset. Both now include SAM 3.1 (1.63 GB) and Qwen3-VL 8B INT8 (8.71 GB, the same file as the Qwen Image 2.1 encoder, so you do not download it twice), sizes as the downloader shows them.
If your ComfyUI reads your SwarmUI Models folder through extra_model_paths.yaml, ComfyUI uses the files downloaded there without a second download.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
23 October 2026 Update V144
This is a Fibo 2.0, MOSS Image Edit and FastH3 Omni update. V144 adds ten new presets: four for Bria's Fibo 2.0, the new open image model that generates and edits images from detailed JSON captions (Dev with the official 50-step recipe and 4-step Turbo, each for text to image and image editing); two MOSS Image Edit presets that turn MOSS-Image into an instruction-based image editor (official 30 steps and Turbo 8 steps); and four presets for FastH3 Omni, MiniMax H3 distilled to 8 steps (References To Video, Text With References To Image, Audio Only and Long Lip Synch). It also adds a Fibo 2.0 prompt guide for ChatGPT, Claude and Gemini, updates the MOSS-Image and MiniMax H3 prompt guides and the two Qwen Image 2.1 presets, and brings the newest ComfyUI with native Fibo 2.0 and FastH3 Omni support plus the matching custom-node and SwarmUI extension updates.
Four native 1920×1088 images from the new Fibo 2.0 Dev Advanced Text To Image preset, with no upscaling: JSON captions, 50 steps, 36 to 38 seconds each on an RTX 5090. Every requested sign, label, menu line and poster line came out exactly as written.
Every result in this post comes from a real run of the shipped presets in ComfyUI started with our Windows_Run_ComfyUI_Default.bat: each preset was opened from the Workflows sidebar and run with the Run button. Only the prompt, the input images or audio, the seed, the output size and the duration were changed; the MOSS Image Edit result in the edit comparison also used a larger reference size, as noted there. The times are ComfyUI's generation times on our RTX 5090; a first run includes loading the models.
Exact V143 → V144 zip audit:
11 new files: the 10 new presets and the Fibo 2.0 LLM prompt guide.
5 changed files: the MOSS-Image and MiniMax H3 prompt guides, the two Qwen Image 2.1 presets and Windows_Install_Or_Update_ComfyUI.bat.
No files were removed. The other 223 files are identical. Fibo 2.0 and FastH3 Omni support arrive through the installers: the ComfyUI update and our updated SECoursesAudioTools, ComfyUI-TeaCache and FoleyExtension.
Here is the complete Presets folder of V144: 106 ready-to-use ComfyUI presets, plus 3 API presets in the api_paid folder. The ten new presets are marked NEW and the two updated Qwen Image 2.1 presets are marked UPDATED:
The presets cover image, video, lip sync, audio and music models: AvatarForever, ERNIE Image, FastH3 Omni, Fibo 2.0, FLUX.1 Dev, Kontext and SRPO, FLUX 2, FLUX 3 Klein, Ideogram 4, Krea 2, LTX 2, LTX 2.3, LTX 2.5, MiniMax H3, MiniMax Music 3, MOSS-Image and MOSS Image Edit, Phantom Wan, Qwen Image, SDXL, Stable Video Infinity 2 Pro, Wan 2.1, Wan 2.2 and Z-Image.
Here are the new and updated presets, with what each one does:
The sections below explain each new model and preset with real results, then the updated Qwen Image 2.1 presets, prompt guides, node packs and the ComfyUI update.
NEW PRESETS — Fibo 2.0: Text To Image and Image Edit, Dev 50 Steps and Turbo 4 Steps
"Fibo 2.0 Dev - Advanced Text To Image - 261022"
"Fibo 2.0 Dev - Advanced Image Edit - 261022"
"Fibo 2.0 Turbo - Text To Image 4 Steps - 261022"
"Fibo 2.0 Turbo - Image Edit 4 Steps - 261022"
Fibo 2.0 is Bria's new open 8B image model. One model generates and edits: text to image, instruction edits, masked edits and edits that combine up to 4 images. It was trained on long structured JSON captions (Bria's VGL format) read by the Qwen3-VL 4B text encoder, and it uses the FLUX 2 VAE. A JSON caption controls every object, its position, the lighting, the camera and each piece of visible text; plain text works too. Our new Fibo 2.0 prompt guide makes ChatGPT, Claude or Gemini write the JSON for you.
Dev and Turbo: Dev is the full model with CFG and gives the best quality with the official recipe (50 steps, CFG 5). Edit Turbo is Bria's distilled editing model (4 steps, CFG 1); Bria lists it for editing only, and in our tests it also makes very good 4-step text to image. Both are INT8 ConvRot files in our Fibo 2.0 Core Bundle.
All four presets share the same easy controls:
One prompt box: a JSON caption or plain text. With no images in the gallery it makes a new image; add 1 to 4 images with + Add images to edit (or drop or paste them) and the prompt becomes the edit instruction, with @image1, @image2 for each image. + Prompt adds more prompts to one run, a folder of .txt files runs a whole prompt list, and Images per prompt makes several images of each prompt.
Output size for text to image: an aspect ratio and megapixels, or width and height. About 1 megapixel is the trained size and 2 megapixels works well.
Edits keep the photo's own size: a 2 MP photo is edited at 2 MP, anything from 1 to 4 MP keeps its size, larger photos are scaled down to 4 MP and smaller ones go to the trained 1 MP size. Choose "Bria default" to edit everything at about 1 MP, as ComfyUI's own node does.
Masked edit: load a photo in 06 OPTIONAL MASKED EDIT, click Paint mask (native editor), paint the area to change and describe what goes there. Every pixel outside the mask is copied back from your photo at the edit size.
Bria's official sampling: the shift follows the image size like Bria's own pipeline (ComfyUI uses the 1 MP value at every size), and Dev text to image uses Bria's JSON negative prompt from the model card; edits leave the negative empty, which ComfyUI encodes as the null caption (a JSON caption with every field empty), like Bria's edit negative.
Prompt syntax: random choices {red|blue} follow the seed in plain-text prompts; weights like (word:1.2) are removed because Qwen3-VL would read them as words; JSON is passed on unchanged (ComfyUI's encoder only drops empty fields, as in training).
Seed fixed or randomized, and an optional LoRA stack, off by default. With Comfy-Org's Fibo 2.0 turbo LoRA (an optional download) in the LoRA stack, Dev also works at 4 steps and CFG 1.
Here is the Dev Advanced Text To Image preset right after the real run of the ramen-stall caption from the top image:
The 1920×1088 image took 37.5 seconds. The first caption asked for glowing shop signs in the alley, and Fibo 2.0 filled them with invented lettering; replacing that phrase with "dark, unlit shop fronts and blank signboards" and "No other readable text anywhere" removed almost all of it, a good example of how exactly Fibo 2.0 follows a JSON caption.
Image edits: here is the same instruction through both Fibo 2.0 edit presets and the new MOSS Image Edit preset: "Change the scene to a snowy winter evening: fresh snow on the roof, the path and the grass, softly falling snow, and warm light glowing in the windows. Keep the building, the TIDE & PAGE BOOKS sign and the chalkboard text exactly the same." The original is the 1920×1088 bookshop from the top image, so the edits kept its 2 MP size:
Fibo 2.0 Dev Advanced Image Edit (50 steps, CFG 5): 87.8 seconds at 2 MP. The closest to the original: the building, the sign, the chalkboard and even the sleeping cat stay, now under fresh snow.
Fibo 2.0 Turbo Image Edit 4 Steps (4 steps, CFG 1): 4.6 seconds at 2 MP, about 19 times faster, with nearly the same result. At 1 MP an edit took 3.4 seconds, including loading the model.
MOSS Image Edit Advanced (30 steps, CFG 4, reference size raised from 1024 to 1920 so that it also edits at 2 MP): 35.6 seconds. It kept the sign and chalkboard text but changed more of the scene around the shop.
NEW PRESETS — MOSS Image Edit: Official 30 Steps and Turbo 8 Steps
"MOSS Image Edit - Advanced Image Edit - 261022"
"MOSS Image Edit - Turbo 8 Steps - 261022"
ChenkinLab's Moss-Image-Edit is a LoRA for MOSS-Image SFT that edits images from a text instruction and up to 4 images. The LoRA alone does not edit: it also needs a reference branch that ChenkinLab ships in its own runtime. Our new MOSS-Image Edit References node adds that branch to ComfyUI's native MOSS-Image model, so these presets run the official recipe with the MOSS-Image models you already have. The newest ComfyUI also has an experimental native version; these presets use our node.
Advanced Image Edit: the official recipe: MOSS-Image SFT with the edit LoRA, 30 steps, Euler / normal, CFG 4, an empty negative and the automatic SFT de-band.
Turbo 8 Steps: adds the 8-step LoRA at 0.7 (8 steps, CFG 1): about 3 seconds per edit, close to the 30-step preset in our season, style and simple background edits. A completely new background behind a busy photo is harder: in a SwarmUI test with the same recipe, Turbo kept a busy street on all 5 seeds where the 30-step recipe made a beach, so start big background changes with Advanced.
How to use: add the image to edit to the gallery and write the instruction: what changes, then what stays. With several images, image 1 is the one edited and the others are references; name them "image 1", "image 2" (or @image1, @image2).
Result size = image 1's size, up to 1024 pixels on the long side; raise reference_max_edge to 1536 for bigger results. The result lines up with image 1 token by token, which is why it keeps the layout and identity.
Same prompt syntax, seed, LoRA stack and folder batch as the MOSS-Image presets of V143.
Here is the Advanced Image Edit preset after a real run with its shipped settings. The image is the presenter portrait from the FastH3 Omni section below, with the instruction "Change her mustard yellow sweater to a dark green raincoat and add round tortoiseshell glasses. Keep her face, hair, smile and the plant-filled studio exactly the same.":
Her face, hair, smile and the plant-filled studio stayed the same; the 688×1024 result took 11.6 seconds.
NEW PRESETS — FastH3 Omni: 8-Step Video, Image, Audio and Long Lip Sync
"FastH3 Omni - References To Video 8 Steps - 261022"
"FastH3 Omni - Text With References To Image 8 Steps - 261022"
"FastH3 Omni - Audio Only With Optional References 8 Steps - 261022"
"FastH3 Omni - Long Lip Synch - Audio + Image 8 Steps - 261022"
FastH3 Omni, by FastVideo and Nuva Lab, is MiniMax H3 Ref2VA distilled to 8 steps: one model makes video with synchronized 32 kHz stereo sound from text plus image, video and audio references. It needs no turbo LoRA; the official recipe (8 steps, Euler / simple, CFG 1, video shift 12, audio shift 3) is already set. It was trained with Video Sparse Attention (VSA), and the video presets run that trained pattern through our updated speed node: sampling about 1.4 times (5-second clip) to 1.9 times (10-second clip) faster than dense attention at 1344×768, with the same quality in our tests.
References To Video: 1344×768, 5 seconds by default (longer works). Add up to 12 references (at most 9 images, 3 videos and 3 audio files) and mention them as @image1, @video1, @audio1. Optional INIT IMAGE = the exact first frame, optional INIT AUDIO = a soundtrack the video must follow (lip sync, timing). Folder batches, merged videos and an optional face refine are included.
Text With References To Image: 1920×1080 stills from text plus up to 9 reference images; the model renders a short frame packet and the first frame is saved as the PNG.
Audio Only: speech, music or sound effects from text, with optional voice, music or video references; saves a FLAC file.
Long Lip Synch: your soundtrack plus a portrait gives one continuous talking video as long as the audio, made in overlapping 8-step windows of about 10 seconds that hold 39 frames at every join. An optional transcript of the exact words and an optional mouth restoration pass are included.
The speed node is on in the two video presets and off in Text With References To Image and Audio Only. A Low VRAM mode, an optional LoRA stack and our RefMods are in every preset.
Here is what FastH3 Omni made from one reference. We made a presenter portrait with Fibo 2.0, used it as @image1 in Text With References To Image and References To Video, made 10 seconds of audio with Audio Only from a text prompt that describes her voice and her line (14.0 seconds), and gave that audio and the portrait to Long Lip Synch:
Text With References To Image: she became a pastry chef in a new scene, with her face, hair and freckles kept: 1920×1080 in 18.9 seconds.
References To Video: a 5-second 1344×768 clip with sound, prompted to show her walking through a greenhouse and speaking to the camera: 56.0 seconds.
Long Lip Synch: the 10-second audio from Audio Only plus the portrait gave a 10-second 832×1248 talking video in 112.0 seconds.
Here are the controls of References To Video, with the portrait as @image1 and the structured prompt our MiniMax H3 prompt guide writes:
Click Add references on the big prompt node, or drag, drop or paste media onto it. Every card shows its token, and typing @ in the prompt inserts one.
The token meter under the prompt estimates how much of the model's context the run uses and which mode it runs in, here "reference to video".
Optional face refine (off by default) regenerates small faces and blends them back; see the fixes below.
And here is Long Lip Synch after the real run with the generated audio and the portrait:
Choose your soundtrack in 1. FULL SOUNDTRACK; the video is as long as the audio. Pick the portrait in 2. OPTIONAL IMAGE (none = audio and text only) and describe the scene in 3. SCENE.
The preset still points to our demo audio and image file names; choose your own files the first time you open it.
On an RTX 5090, 15 seconds of audio (two windows) at 832×1248 take about 3.5 minutes. Smaller windows (141 or 192 frames) use less VRAM.
UPDATED PRESETS — Qwen Image 2.1: Text Encoder File Name
"Qwen Image 2.1 - Unified Generate Edit Inpaint RGBA - 260920"
"Qwen Image 2.1 - Unified Generate Edit Inpaint RGBA Ultimate Quality - Slow - 260922"
Both presets now load the Qwen Image 2.1 text encoder as qwen3vl_8b_int8_convrot.safetensors, its original name. Our downloader used to save it as qwen3vl_8b.safetensors, the name SwarmUI uses for Ideogram 4's different encoder, so SwarmUI downloaded 9.35 GB again for Qwen Image 2.1. The updated downloader (SwarmUI_Model_Downloader_v184.zip) saves it under the original name, and both presets were tested with it.
If you downloaded Qwen Image 2.1 or Ideogram 4 with an older downloader: it saved all three Qwen Image 2.1 encoders and both Ideogram 4 encoders as text_encoders/qwen3vl_8b.safetensors, so the last download won. Check that file's size (SwarmUI may have added about 64 KB of metadata to a file it used):
9,350,798,360 bytes: the Qwen Image 2.1 INT8 ConvRot encoder; rename it to qwen3vl_8b_int8_convrot.safetensors.
17,534,334,616 bytes: the BF16 encoder; rename it to qwen3vl_8b_bf16.safetensors.
6,312,105,364 bytes: the W4A8 encoder; rename it to qwen3vl_8b_w4a8.safetensors.
10,588,637,512 or 6,305,221,764 bytes: Ideogram 4's own FP8 or NVFP4 encoder; keep it as it is.
After a rename, Ideogram 4 gets its own encoder again from its core bundle or SwarmUI's automatic download. Or download the Qwen Image 2.1 Core Bundle again: only the 9.35 GB text encoder (shown as 8.71 GB in the downloader) is downloaded, the model and VAE are kept. The Qwen presets do not download models themselves, so they need the renamed or re-downloaded file.
NEW AND UPDATED PROMPT GUIDES for ChatGPT, Claude and Gemini
NEW — Fibo_2.0_Enchance_Prompt_Feed_For_LLMs.txt: give it to an LLM with your idea (attach the image for an edit) and it writes a Fibo 2.0 JSON caption in Bria's format: objects, background, lighting, camera, style and every visible text in text_render, for text to image, single-image edits, edits with 2 to 4 images (<image_1>, <image_2>) and masked edits.
UPDATED — MOSS_Image_Enchance_Prompt_Feed_For_LLMs.txt: a new image edit part: one short instruction that says the change first and then what stays, images named "image 1" and "image 2", and an empty negative prompt.
UPDATED — MiniMax_H3_Enchance_Prompt_Feed_For_LLMs.txt: a new FastH3 Omni section. Prompt FastH3 Omni exactly like MiniMax H3 Ref2VA; the presets also take an init image and a locked init audio, and steps or samplers never go into the prompt.
FIXES FROM OUR V144 FEATURE TESTS
We ran all ten new presets on the ComfyUI our installer gives you today. These problems were found and fixed before this release:
Fibo 2.0 Dev text to image on the newest ComfyUI: ComfyUI now puts Fibo 2.0's negative prompt into Bria's chat template itself, and our node did the same, so the negative was wrapped twice. Our node no longer wraps it. On the newest ComfyUI the four Fibo 2.0 presets now give the same pixels, seed for seed, as before this ComfyUI change.
FastH3 Omni optional face refine: the face-pass instruction asked for "beard strands", which gave a woman a beard in our test; the words are gone from the presets and from the node's default. With FastH3 the old constant 0.55 pass was rejected by our hallucination guard on small faces and changed nothing, so the FastH3 presets now use 0.35, which repaired them (the image preset also uses FastH3's 8 steps, Euler / simple, for the face pass).
FastH3 Omni model download links: the model loader pointed ComfyUI's missing-model download at MiniMax H3 Ref2VA and the text encoder loader at the NVFP4 encoder; they now point to the FastH3 Omni INT8 model and the INT8 Qwen3-VL 32B encoder the presets load.
Token meter: in presets with RefMods, the meter under the prompt ignored the gallery's references and showed "text to video" for a reference run; it now counts them.
Long Lip Sync without a soundtrack now stops with a clear message instead of a Python error.
Notes and titles in the FastH3 presets now match their settings (speed mode, face refine, seed).
HOW TO UPDATE
Download the latest ComfyUI_V144.zip from this post.
Extract it over your existing installer folder and overwrite the older files.
Close ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat. It updates ComfyUI to the newest version with Fibo 2.0 and FastH3 Omni support, and our ComfyUI-TeaCache and FoleyExtension, plus ComfyUI-Manager. The Fibo 2.0 and FastH3 Omni presets do not run on older ComfyUI.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install/update bundle 100. It updates SECoursesAudioTools with the Fibo 2.0 and MOSS-Image Edit nodes and ComfyUI-KJNodes, and keeps SwarmUI ExtraNodes and ComfyUI-Easy-Use, which the new presets also use, up to date.
Restart ComfyUI and refresh the browser tab.
Use our latest SwarmUI Model Downloader (SwarmUI_Model_Downloader_v184.zip or newer) and download the bundles from the SwarmUI Bundles tab (the downloader shows every size in binary GB; the bundle totals below are its numbers, the file sizes are decimal GB):
FastH3 Omni Core Bundle (first card, 49.78 GB, 9 files): the FastH3 Omni 8-step INT8 ConvRot model (22.2 GB), the INT8 Qwen3-VL 32B text and vision encoder (27.1 GB), the MiniMax H3 video and audio VAEs and the face detection, recognition and restoration models of the optional face refine and mouth pass. BF16, FP8 and NVFP4 versions of the model and the NVFP4 and BF16 encoders are optional downloads. If you downloaded the MiniMax H3 bundle with V183, its INT8 video VAE was Kijai's file (3.17 GB) saved under the name of Comfy-Org's file that V184 expects, so the downloader downloads Comfy-Org's file once (2.81 GB, shown as 2.62 GB).
Fibo 2.0 Core Bundle (second card, 20.50 GB, 4 files): the Dev and Edit Turbo INT8 ConvRot models (8.4 GB each), the INT8 Qwen3-VL 4B encoder (4.9 GB) and the FLUX 2 VAE. The BF16 files and Comfy-Org's turbo LoRA for Dev are optional downloads.
MOSS-Image Core Bundle (third card, 21.91 GB, 7 files): it now also has MOSS_Image_Edit_LoRA.safetensors (0.30 GB), ChenkinLab's edit LoRA, which the MOSS Image Edit presets use.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
18 October 2026 Update V143
17 October 2026 Massively Detailed Windows Tutorial : https://youtu.be/yOj9PYq3XYM
Presets Folder Content In Latest Zip File
This is a MOSS-Image update. V143 adds six new presets for MOSS-Image, the new open text-to-image model from OpenMOSS: Text To Image with the official 50-step quality recipe, Fast 25 Steps, Turbo 8 Steps, Layout Boxes, and Base model versions of Text To Image and Layout Boxes. It also adds a MOSS-Image prompt guide for ChatGPT, Claude and Gemini, a ComfyUI notebook for Google Colab, an updated AvatarForever preset with a live preview panel, fixes for the VRAM-optimized launcher and the installer, the newest ComfyUI with native MOSS-Image support, and the matching custom-node and SwarmUI extension updates.
Four native 1920×1088 images from the new Advanced Text To Image preset, with no upscaling: 50 steps, 32 to 41 seconds each on an RTX 5090. Every label and poster line came out exactly as written, and the chalkboard is spelled correctly.
Every result in this post comes from a real run of the new presets with their default settings; only the prompt (for the layout, also the boxes), seed and output size were changed. The timings were measured on our RTX 5090 after the models were loaded.
Here is the complete Presets folder of V143: 96 ready-to-use ComfyUI presets, plus 3 API presets in the api_paid folder. The six new MOSS-Image presets are marked NEW and the updated AvatarForever preset is marked UPDATED:
The presets cover image, video, lip sync, audio and music models: AvatarForever, ERNIE Image, FLUX.1 Dev, Kontext and SRPO, FLUX 2, FLUX 3 Klein, Ideogram 4, Krea 2, LTX 2, LTX 2.3, LTX 2.5, MiniMax H3, MiniMax Music 3, MOSS-Image, Phantom Wan, Qwen Image, SDXL, Stable Video Infinity 2 Pro, Wan 2.1, Wan 2.2 and Z-Image.
NEW PRESETS — MOSS-Image: Text To Image, Fast, Turbo, Layout Boxes and Base
"MOSS Image - Advanced Text To Image - 261017"
"MOSS Image - Fast 25 Steps - 261017"
"MOSS Image - Turbo 8 Steps - 261017"
"MOSS Image - Advanced Layout Boxes - 261017"
"MOSS Image Base - Advanced Text To Image - 261018"
"MOSS Image Base - Advanced Layout Boxes - 261018"
Here are the new and updated presets, with what each one does:
MOSS-Image is a new open image model from OpenMOSS: a 5B diffusion transformer that reads your prompt with the Qwen3.5-9B language model and uses the FLUX 2 VAE. It renders clear, accurate text inside images, accepts long and detailed prompts of up to about 4,000 tokens, and, as an experimental feature of the model, can place objects, characters and text in boxes you draw. It makes images from text; it does not edit images or use reference images.
SFT and Base: SFT is the official fine-tune of Base on curated data, and the four main presets use it; in the official comparison it groups text more clearly. Base is the raw model without post-training, with more varied images, meant for LoRA training; two Base presets are included.
All six presets share the same easy controls:
One prompt box: plain language or tags, and in the Text To Image presets also a JSON layout. Long, detailed prompts work best. Put exact text for signs, labels and posters in quotes, and use real line breaks to control the lines.
Output size: choose an aspect ratio and megapixels, or type the width and height. Any aspect ratio works; 1 to 2 megapixels is recommended. Sizes snap to multiples of 16.
Batches: + Prompt adds more prompts to one run, a folder of .txt files runs a whole prompt list (JSON layouts too), and Images per prompt makes several images of each prompt.
Prompt syntax: random choices {red|blue} follow the seed in the main prompt, added prompts and folder prompts, so a fixed seed repeats the same prompt. Schedules [red:blue:0.5] and alternation [red|blue] work in plain-text prompts of the Text To Image presets. Weights like (word:1.2) are removed because MOSS-Image's text encoder would read them as words; describe emphasis in words instead.
JSON layouts in the Text To Image presets: a complete MOSS-Image layout is sent unchanged, only without extra spaces. Pixel-based layout drafts from the official MOSS prompt enhancer or our prompt guide are converted to the model's format automatically and scaled to your output size. In the Layout Boxes presets, the prompt box holds only the caption of the whole image.
Automatic SFT de-band: the SFT model can draw faint vertical color stripes on skies, fog and plain walls. The presets apply a community fix to SFT automatically and skip it for Base.
Seed: fixed, or randomized on every run, with an optional variation seed and strength.
Optional LoRA stack for MOSS-Image LoRAs, off by default.
Attention set to PyTorch: it matched the official MOSS-Image images in our tests, and it also prevents black images with the BF16 files when ComfyUI starts with Sage attention, as our launchers do (RTX 3090 test). Keep it on PyTorch in your own BF16 workflows.
The model, text encoder, VAE and latent are already connected. To use the BF16 files, switch the model loader. Use Base only in the 50-step presets: the Fast and Turbo recipes were made for SFT.
Here is the Advanced Text To Image preset after a real run of a magazine cover prompt:
All five text lines came out exactly as written, from the huge WANDER title to the small issue line. The 1024×1536 image took 21.8 seconds with the official 50-step recipe.
Fast 25 Steps, Turbo 8 Steps and Base: here is the same prompt and seed through all four text-to-image presets at 1920×1088:
Advanced Text To Image (SFT, 50 steps, Euler / normal, CFG 3, shift 5, the official recipe): the most faithful results, especially for text and designs. 34.3 seconds here, 12.9 seconds at 1024×1024.
Fast 25 Steps (ER-SDE / simple, CFG 3, Temporal Score Rescaling 0.95, layer skip 22 and 24 from 10% to 70% of the steps): 16.6 seconds here, about half the time of the 50-step preset. Large text holds up well; very small text can occasionally lose a letter.
Turbo 8 Steps (the community 8-step LoRA by lodestones at 0.7, Euler / simple, CFG 1): 3.4 seconds here, about a tenth of the 50-step time. The 8-step LoRA is a work in progress, so its images differ more from the 50-step ones and can show extra signs.
Base Advanced Text To Image (Base model, the official 50-step recipe): more varied images; 32.5 seconds here. For text-heavy designs, start with SFT: in the official Base versus SFT example, SFT groups text more clearly.
Layout Boxes: draw a box for each object, character or line of text and describe it. The main prompt becomes the caption of the whole image:
Each box is an object or exact text, with its own description. To turn object boxes into characters, type their numbers in character_boxes, here 8 for the barista.
The canvas follows the output size, and the preset builds MOSS-Image's layout JSON from your boxes.
The box editor needs Settings → Modern Node Design (Nodes 2.0). Turn it on once.
Boxes guide placement. They are not masks, so positions can vary a little.
Here is the result of that layout in the same preset:
Every box landed where it was drawn: the headline, three drinks with their exact names and prices, and the waving barista as a character. The 1536×1024 image took 22.6 seconds.
UPDATED PRESET — AvatarForever: Live Preview Panel and Tested Mouth-Movement Settings
"AvatarForever Unified Image or Text + Audio To Long Video - 260920"
New Live Preview & Final Video panel to the right of the generation controls. It shows a frame preview after each sampling step, starting with the first, then plays the finished MP4.
New tested preset settings: a 512×768 canvas and 16 latent frames per chunk, instead of 832×1248 and 4. In our singing demo, 4-frame chunks kept the lips closed, while 16-frame chunks gave sustained mouth movement through the whole 18-second song. At 1024×1024 the mouth stayed mostly closed even with 16-frame chunks, so check a short sample when you change the image, audio, resolution or chunk size. Larger chunks need more memory; 4-frame chunks still work for speech and use less.
Sharper previews: the live preview uses the small taeltx2_3 preview decoder when it is installed. It is now in the LTX 2.3 Avatar Forever Core Bundle of our SwarmUI Model Downloader V183. Without it, the preview uses a coarser latent view.
Your own AvatarForever workflows keep working and now show live previews on the sampler. Set live_preview to off to turn them off.
NEW PROMPT GUIDE — MOSS-Image Feed for ChatGPT, Claude and Gemini
A new MOSS_Image_Enchance_Prompt_Feed_For_LLMs.txt is included. Give it to ChatGPT, Claude or Gemini together with your request, and it writes prompts that MOSS-Image understands:
Detailed captions for photos, illustrations and designs, with the exact visible text in quotes and where and how it appears.
Layouts as pixel-based JSON drafts for the Text To Image presets, which convert them to MOSS-Image's own format automatically, or a box-by-box plan to draw in the Layout Boxes editor.
Tag-style prompts for anime, in the recommended tag order.
The right settings: output size, CFG for short prompts, the negative prompt and when to use SFT or Base.
NEW — ComfyUI Colab Master Notebook for Google Colab
"ComfyUI_Colab_Master_Notebook.ipynb"
A new Google Colab notebook runs ComfyUI with ComfyUI-Manager on a Colab GPU and opens it in your browser through a Cloudflare link. It sets up Qwen-Image 2.1 for you:
Install: ComfyUI (the newest stable release, or master if you choose it) with ComfyUI-Manager and cloudflared, which makes the link without an account. Colab's own PyTorch is kept.
Qwen-Image 2.1 in the right size: the official ComfyUI files are picked from your GPU memory and system RAM: INT8 with the W4A8 text encoder (14.2 GB) when either is below 20 GiB, as on a T4; INT8 with the INT8 text encoder (17.3 GB), the official template's files, when both have at least 20 GiB; or full BF16 (32.4 GB) when you choose it. Broken downloads resume when you run the cell again.
Any other model: paste up to four direct download links, such as LoRAs or upscalers, and choose the folder or let the notebook read it from the link.
Start: you get a trycloudflare.com link, and ComfyUI opens with the Qwen-Image 2.1 text-to-image workflow already loaded, with the official settings (Euler, simple, 25 steps, CFG 1, 1024×1024).
Keep your work: one cell packs your outputs and saved workflows into a ZIP to download. The notebook also explains how to upload images and videos, and what is kept when the runtime ends.
Open Google Colab, choose Upload notebook, select the file, pick a GPU runtime (T4 or better) and run steps 1 to 4 in order. T4 GPUs can be allocated on free accounts, but availability is not guaranteed, and Colab says free runtimes used mainly through an external web UI may be terminated. The Cloudflare link is public, so don't share it.
LAUNCHER AND INSTALLER FIXES
Windows_Run_VRAM_Optimized.bat: option 3 (normal VRAM), and VRAM choice 3 in Custom mode, passed --normalvram, which ComfyUI removed in version 0.22.0, so ComfyUI stopped at once with "unrecognized arguments: --normalvram". Both now start ComfyUI with its default memory management and no extra flag.
Windows_Run_VRAM_Optimized.bat also describes the memory modes as current ComfyUI uses them. LOWVRAM's --lowvram has no effect while dynamic VRAM is on, the default on NVIDIA GPUs, and otherwise only runs the text encoders on the CPU; NOVRAM turns dynamic VRAM off. Pressing Enter in the cache menu now keeps ComfyUI's own default, RAM-pressure caching, instead of the older classic caching; choose 1 for classic caching.
Windows_Install_Or_Update_ComfyUI.bat: the pip update and the uv install now ignore your personal pip settings (the pip.ini in your user profile and PIP_ environment variables), so a custom package index set there for another project can no longer stop the installer at this step. Its Windows requirements tutorial link now points to the current video: https://youtu.be/ptuLL0C72Ms
HOW TO UPDATE
Download the latest ComfyUI_V143.zip from this post.
Extract it over your existing installer folder and overwrite the older files.
Close ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat. It updates ComfyUI to the newest version, with MOSS-Image support. The MOSS-Image presets do not run on older ComfyUI.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install/update bundle 100. It updates SECoursesAudioTools with the new MOSS-Image nodes and the AvatarForever Live Preview & Final Video node that the updated AvatarForever preset uses, and keeps FoleyExtension, SwarmUI ExtraNodes and ComfyUI-Easy-Use, which the MOSS-Image presets also use, up to date.
Restart ComfyUI and refresh the browser tab.
For the Layout Boxes editor, turn on Settings → Modern Node Design (Nodes 2.0).
Use our latest SwarmUI Model Downloader (SwarmUI_Model_Downloader_v183.zip or newer) and download the MOSS-Image Core Bundle, the first card of the SwarmUI Bundles tab. It has the six files the six presets use, about 23.2 GB in total (the downloader counts in GiB, so it shows 21.64 GB):
moss_image_sft_int8_convrot.safetensors (5.41 GB), the SFT model in INT8 ConvRot.
moss_image_base_int8_convrot.safetensors (5.41 GB), the Base model in INT8 ConvRot.
qwen3.5_9b_int8_convrot.safetensors (9.75 GB), the Qwen3.5-9B text encoder in INT8 ConvRot.
Flux/flux2-vae.safetensors (0.34 GB), the FLUX 2 VAE, the same file the FLUX 2 and FLUX 3 presets use.
MOSS_Image_SFT_DeBand_L37-41.safetensors (1.13 GB), the SFT de-band fix.
MOSS_Image_SFT_8_Steps_DMD_LoRA.safetensors (1.19 GB), the 8-step LoRA of the Turbo preset.
The INT8 ConvRot model files are about half the size of BF16, and in our RTX 3090 test they were about twice as fast (48 versus 99 seconds at 50 steps). The BF16 models, a smaller W4A8 text encoder (select it in the text encoder loader), the small-decoder FLUX 2 VAE and the extra de-band fix for layers 32–36 are optional downloads under Image Generation Models → MOSS-Image Models. The de-band author recommends the extra fix for images 1536 pixels wide or larger and for big, smooth skies; select it in the de-band node.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
17 October 2026 Update V142
This is a FLUX 3 Klein update. V141 adds three new presets for Black Forest Labs' new FLUX 3 Image Klein model: Text To Image, Image Edit with up to 10 reference images, and Layout Boxes. It also adds a FLUX 3 Klein prompt guide for ChatGPT, Claude and Gemini, our new HQ INT8 ConvRot FLUX 3 Klein model, ComfyUI 0.41.0 with native FLUX 3 support, and the matching custom-node and SwarmUI extension updates.
A native 3840×2160 image from the new Text To Image preset, with no upscaling: 16 steps in 23 seconds on an RTX 5090.
Every result in this post comes from a real run of the new presets with their default 16 steps, Euler / simple, CFG 1. The timings were measured on our RTX 5090 after the models were loaded.
NEW PRESETS — FLUX 3 Klein: Text To Image, Image Edit and Layout Boxes
"FLUX 3 Klein - Advanced Text To Image - 261015"
"FLUX 3 Klein - Advanced Image Edit - 261015"
"FLUX 3 Klein - Advanced Layout Boxes - 261015"
FLUX 3 Image Klein is a 5.6B image model with the Qwen3-VL-4B text encoder. One model does text to image, image editing, up to 10 reference images, layout boxes and text rendering. Our presets run the fast distilled model with its recommended 16 steps and no guidance (CFG 1), using Euler / simple.
All three presets share the same easy controls:
One prompt box with an image gallery: write the whole scene. Add up to 10 images and refer to them as @image1 … @image10 (@img1, @picture1 and @pic1 also work). With no image, it is text to image.
Output size: choose an aspect ratio and megapixels, or type the width and height. Sizes snap to multiples of 16.
Batches: + Prompt adds more prompts to one run, a folder of .txt files runs a whole prompt list, and Images per prompt makes several images of each prompt.
Advanced prompt syntax: weights (ceramic:1.1), random choices {red|blue}, schedules [red:blue:0.5], alternation [red|blue] and <break>. Random choices follow the seed, so a fixed seed repeats the same prompt.
Seed: fixed, or randomized on every run, with an optional variation seed and strength.
Reference size: one Reference megapixels setting resizes every reference image and keeps its shape (1 MP by default, up to 4 MP). It is separate from the output size.
Optional LoRA stack for FLUX 3 LoRAs, off by default.
The model, text encoder, VAE and latent are already connected. There is nothing else to set up.
Here is the Text To Image preset after a real run of a poster prompt with English and Turkish text:
All four text phrases came out exactly as written, including the Turkish characters. The 1024×1536 image took 3.2 seconds. The first run takes longer because it loads the models.
Image Edit: the same workflow with your image in the gallery. Here it turns a celadon museum vase cobalt blue and keeps its shape, handles, framing and shadow:
Say what changes and what stays, and refer to your photo as @image1.
Match the output aspect ratio to your photo to keep its framing. This 1024×1536 edit took 5.2 seconds.
The preset opens with a sample card. Add your own image with + Add reference images and remove any card marked "file missing".
Up to 10 reference images in one prompt: here ten separate museum photos became one display wall, each object in its requested niche:
All ten objects appear in the requested order and closely match their photos. The 1680×944 image took 41 seconds with all ten references at 1 MP each. The reference photos are public domain (CC0) images from The Cleveland Museum of Art.
Layout Boxes: draw a box for each object or line of text, describe it, and refer to it in the caption as <obj_1>, <obj_2> and so on:
Each box is an object or exact text, with its own description. The preset builds FLUX 3's layout table from your boxes, and the boxes follow the output size.
The box editor needs Settings → Modern Node Design (Nodes 2.0). Turn it on once.
Boxes guide placement. They are not masks, so positions can vary a little.
Here is the result of that layout in the same preset:
All four terrariums and captions landed where their boxes are, and the four captions are spelled exactly. The 1024×1024 image took 2.5 seconds.
NEW MODEL — Our FLUX 3 Klein HQ INT8 ConvRot
"FLUX-3-Image-Klein-INT8-ConvRot-HQ.safetensors"
The presets use our own INT8 ConvRot conversion of the distilled FLUX 3 Klein model. On the updated ComfyUI it loads with the normal Load Diffusion Model node, so no special loader is needed.
6.07 GB instead of 11.1 GB for the original BF16 model. While denoising at up to 1024×1024, the model alone peaked at about 6 GiB of VRAM instead of about 10.6 GiB for BF16 (RTX 3090 test).
About 2× faster per step than BF16 (1.9–2.3× in our RTX 3090 tests, model only).
Closer to BF16 than the official INT8 file: in our test of 40 sampling states, its mean output error against BF16 was 2.14%, versus 3.05% for the official INT8 ConvRot file with its default settings and 2.27% with matched attention. One held-out editing case favored the official file with matched attention. In exchange, our file is 0.25 GB larger and was 6–22% slower per step in the same RTX 3090 test. These are numerical measurements made with the BF16 text encoder, not image-quality percentages.
The first text block and the last joint block, the most sensitive in our tests, stay in BF16, as do the small conditioning, normalization and output layers. The other 164 linear layers use native INT8 ConvRot.
NEW PROMPT GUIDE — FLUX 3 Klein Feed for ChatGPT, Claude and Gemini
A new FLUX_3_Klein_Enchance_Prompt_Feed_For_LLMs.txt is included. Give it to ChatGPT, Claude or Gemini together with your request and images, and it writes prompts that FLUX 3 Klein understands:
Complete scene captions for text to image, with the most important details first.
Edit prompts that say what changes and what must stay, and multi-reference prompts that give every image a clear role.
Exact text for posters and designs, in the original language and spelling.
Layout prompts in the FLUX 3 box format, with the <obj_1> labels these presets use.
The right settings and the ComfyUI / SwarmUI details, such as @image1 meaning FLUX 3's first reference image.
AvatarForever Preset Now Has Live Preview Feature
2:3 and 3:2 aspect ratios works best - 1:1 not that good
HOW TO UPDATE
Download the latest ComfyUI_V141.zip from this post.
Extract it over your existing installer folder and overwrite the older files.
Close ComfyUI, then run Windows_Install_Or_Update_ComfyUI.bat. It updates ComfyUI to the newest version, with FLUX 3 support. The FLUX 3 presets do not run on older ComfyUI.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install/update bundle 100. It updates SECoursesAudioTools with the new FLUX 3 nodes and keeps FoleyExtension, SwarmUI ExtraNodes and ComfyUI-Easy-Use, which the presets also use, up to date.
Restart ComfyUI and refresh the browser tab.
For the Layout Boxes editor, turn on Settings → Modern Node Design (Nodes 2.0).
Use our latest SwarmUI Model Downloader (SwarmUI_Model_Downloader_v181.zip or newer) and download the FLUX 3 Klein Core Bundle, the first card of the SwarmUI Bundles tab. It has the three files all three presets use, about 11.7 GB in total (the downloader counts in GiB, so it shows 10.85 GB):
FLUX-3-Image-Klein-INT8-ConvRot-HQ.safetensors (6.07 GB), our HQ INT8 ConvRot model.
qwen3vl_4b.safetensors (5.24 GB), the FP8 Qwen3-VL-4B text encoder.
Flux/flux2-vae.safetensors (0.34 GB), the FLUX 2 VAE, which FLUX 3 uses too.
The text encoder is the same file Krea 2 uses and the VAE is the FLUX 2 VAE, so if you already have those, only the 6.07 GB model is new.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
10 October 2026 Update V140
ComfyUI V140 is ready. Here is everything we changed since V138: 4 new MiniMax H3 presets, optional MiniMax H3 RefMods (reusable reference files) in our MiniMax H3 presets, a native sparse attention engine in our MiniMax H3 Speed Optimizer, fixes that keep our nodes working on the newest ComfyUI (tested on ComfyUI 0.39.0), and fixes for every preset that failed when we ran all presets with real generations.
New MiniMax H3 Presets
MiniMax H3 Lip Synch - Turbo 4 Steps (672x1216 and 832x1248)
Here is the 672x1216 preset after a real run with its demo audio and portrait:
Fast talking videos from your audio: upload an audio file, optionally add an image, describe the scene and run. The video length follows your audio. Without an image it works as audio + text. With a video in the reference gallery, it makes a new performance from that video.
Turbo 4 steps by default with the LightX2V FL2VA Turbo 4-step v1.2 LoRA, at 24 fps. A 15-second video took about 165 seconds at 672x1216 on an RTX 5090. Set "4. QUALITY / SPEED / VRAM" to FULL QUALITY for the 50-step path.
Two portrait resolutions: 672x1216, or 832x1248 for more detail. These are 4-step portrait versions of our existing MiniMax H3 Lip Synch preset (1024x1024, 8 steps).
The 832x1248 preset saves the original video first, then runs the new CodeFormer mouth pass:
The 832x1248 preset adds a CodeFormer mouth pass (on by default). It restores only the mouth area of the finished video and saves it as a second MP4, next to the original video.
Also included: our optional face refinement (off by default) and the new optional RefMods.
Tip: for your own audio, put its exact words in the <d>[English] ... </d> part of the scene prompt, or clear the scene prompt to use the automatic one. 5-15 seconds per generation is recommended; longer audio still works in one pass.
MiniMax H3 Long Lip Synch - Native Latent Continuation - Full Audio
Here is the Long Lip Synch preset with its default settings:
One full soundtrack in, one continuous video out: give it a complete audio file plus a portrait image or gallery references. It makes one continuous 24 fps lip-synced video as long as your audio. There is no duration setting and no cutting.
How it stays continuous: it generates overlapping windows of about 10 seconds (243 frames) and carries the last 1.6 seconds (39 frames) of each window forward as clean latents, so nothing is re-encoded between windows. The audio is encoded once, and your original soundtrack is put back into the final MP4.
Works with references and RefMods in every window, unlike the TaoMate streaming preset. The first-frame image anchors only the start.
Defaults: 832x1248 and Turbo 4 steps (FL2VA 4-step v1.2, or Ref2VA 4-step v0.1 when you use references). Use 512x768 for faster drafts. SOL sparse attention and FirstBlockCache are separate switches, both off by default.
Outputs: the final MP4 with your original audio, plus a JSON report with the timing of every window. On an RTX 5090, 30 seconds of 512x768 video took 105 seconds with normal attention.
Experimental: very long videos can change framing over time. In our 15-clip test, 13 clips passed; the 45 and 60-second clips at 832x1248 drifted into close-ups.
MiniMax H3 RefMod Builder - Images, Video, Audio
Add your source media, name your RefMod and run. Only the MiniMax H3 VAEs are used:
Build your own RefMod in one run: add photos of one subject, a video (its soundtrack is included), audio, or a mix. Name it and run. Only the MiniMax H3 VAEs run: no diffusion model, no text encoder and no training.
Options: source resolution (512 by default), "one character" (stacks the photos into one reference) or "separate references" (each photo gets its own picture label), and optional compression (32, 16 or 8) for fewer tokens at the cost of detail.
Ready to use: files are saved to ComfyUI/models/refmods and existing files are never overwritten. Pick the file in any MiniMax H3 preset's Optional RefMods node and write @refmod1 in your prompt.
MiniMax H3 RefMods: Reusable References
A RefMod is a small .safetensors file that stores already-encoded MiniMax H3 references: images, video and/or audio of a character, scene or voice. Build it once, then reuse or share it in any MiniMax H3 preset without encoding the source media again. It is not a LoRA and needs no training.
Here is the new Optional RefMods node with an imported RefMod, used as @refmod1 in the prompt:
New "MiniMax H3 Optional RefMods" node in all 9 existing MiniMax H3 generation presets (Text To Video, Image or Video To Video, References To Video, Audio Only, Text With References To Image and Lip Synch) and in the new presets. It is empty by default, so your results do not change unless you add a RefMod.
Easy picker: "+ Add RefMod", "Import files" and "Refresh" buttons. Each row has an on/off box, the file (with its contents, for example "image + audio"), a strength from 0 to 1, a visual/audio filter and up/down ordering. Imported files go to ComfyUI/models/refmods and are checked before use.
Use them in your prompt with @refmod1, @refmod2 and so on. They are numbered after your gallery images, videos and audio.
Strength sets how much reference detail is kept. It is not a LoRA weight.
RefMods alone are enough: the reference mode and auto mode nodes count RefMods as references, so a RefMod without any gallery image switches generation to the references (Ref2VA) path.
Compatible files: RefMods made with ComfyUI-MiniMaxH3Mod and the Fantastic MiniMaxH3 PromptBuilder load too, without installing those node packs.
TaoMate streaming does not use RefMods; its preset note now points you to the new Long Lip Synch preset for RefMods with long audio.
MiniMax H3 Speed Optimizer: Native Sparse Attention
Here is the MiniMax H3 Speed Optimizer node, shown in the Text With References To Image preset, where it is off by default:
SOL sparse attention now runs on ComfyUI's own MiniMax H3 sparse attention (our ComfyUI-TeaCache fork). The old bundled attention kernels are gone, so there is nothing extra to compile.
Faster than our old SOL: the old version repeated work and was slower than normal attention at 832x1248 (207 s against 116 s). With the new default protected SOL settings, a 10-second 832x1248 Turbo 4-step video took 99 seconds instead of 111 seconds with normal attention (1.12x, RTX 5090, FirstBlockCache off for this test). At 512x768 the gain was 1.06x. Sparse attention and caching are approximations, so check the quality of your own results.
Auto mode measures before it switches: it uses sparse attention only when it is at least 5% faster. The Text To Video and Image or Video To Video presets have the optimizer on in auto mode, so they now use the native path automatically when it pays off.
FirstBlockCache works together with other speed-ups: it now keeps patches that are already on the model, such as the sparse attention added by the lip-sync Turbo node, instead of replacing them.
Old presets load unchanged. The main switch is now called "H3 optimizations" instead of "4x speedup", and "native" is the new default backend.
Other MiniMax H3 Improvements
Correct colours in streamed MiniMax H3 videos: TaoMate streaming and long lip-sync videos are now tagged with the colour format they are encoded in (BT.601 limited range), so players read the colours correctly.
No torchaudio needed for MiniMax H3 audio: init audio, streaming and long lip sync now resample any soundtrack with ComfyUI's own resampler, with bit-identical results (details below).
Works With the Latest ComfyUI
ComfyUI removed torchaudio and changed its attention and Linear layer code. We updated our nodes so everything keeps working on the newest ComfyUI:
No more torchaudio needed: our MiniMax H3 streaming and long lip-sync nodes (SECoursesAudioTools), the init audio nodes (FoleyExtension) and the audio models in our ComfyUI-WanVideoWrapper fork (HuMo, Ovi, FantasyTalking, MultiTalk and LongCat) now resample audio with ComfyUI's own resampler. Before, on installs without torchaudio, they failed on any soundtrack that was not already at the model's sample rate, and HuMo and Ovi did not load at all. The output is bit-identical: 88 test comparisons had zero difference, and a real MiniMax H3 Turbo 4 video was pixel-identical on all 120 frames.
Here are the new torchaudio-free Mel-Band RoFormer nodes inside the LTX2.3 audio lip-sync preset:
New torchaudio-free Mel-Band RoFormer vocal separation nodes (SECoursesAudioTools: "SE Mel-Band RoFormer Model Loader" and "SE Mel-Band RoFormer Sampler"). They load the same checkpoint and give bit-identical vocals, so the LTX2 and LTX2.3 audio lip-sync presets keep their vocal separation even where the original Mel-Band RoFormer node pack cannot load.
GGUF models work again (our ComfyUI-GGUF fork): on the latest ComfyUI every quantized GGUF model failed with "unexpected keyword argument 'input_act'". Fixed and verified bit-exact on 57 test cases.
FLUX 2 Mistral GGUF text encoder loads again (ComfyUI-GGUF): the GGUF Tekken tokenizer now gives ComfyUI the special token count it needs. Token IDs match the official tokenizer.
Much faster FLUX 2 GGUF on limited VRAM (ComfyUI-GGUF): GGUF text encoders are now fully unloaded when VRAM is needed, instead of leaving 6.6 GB in VRAM. FLUX 2 Low RAM sampling went from 200-309 s/it to 1.75 s/it on a 32 GB GPU.
LTX 2.3 Foley presets fixed: ComfyUI-LTXVideo is now installed from our fork, which fixes the STG error "'AttentionTensorContainer' object has no attribute 'shape'". Existing installs switch to the fork automatically on their next update.
ControlNet Aux keeps the MIT albumentations package (our comfyui_controlnet_aux fork): upstream switched to AlbumentationsX, which is AGPL-3.0, sends usage telemetry by default and conflicts with insightface (used by ReActor).
Preset Fixes and Updates
We ran every preset with real generations on the latest ComfyUI and fixed every one that failed. The FLUX 2 Low RAM presets now free VRAM between prompt encoding and sampling:
FLUX 2 - 8 Steps Fast HQ / Realism - Low RAM: two new "Free VRAM before sampling" nodes unload the models after the prompts are encoded. Speed went from 200-309 s/it to 1.75 s/it, 35-75 seconds per image on a 32 GB GPU. The Realism Low RAM preset also had the HQ preset's guidance of 7; it is now 1, like the normal Realism preset.
Krea 2 Raw Base, Krea 2 Turbo and Krea 2 Turbo Image Edit: the saved sampler values were shifted by one position (for example steps became 1 and CFG became "euler"). Fixed.
LTX2 Image To Video and LTX2 Text To Video 8 Steps: the seed was empty, so generation failed. Fixed.
LTX2.3 Extend Any Video With Audio: the input video is now resized to a multiple of 64, so the source and the half-resolution second stage always match (1280x720 is unchanged).
LTX2.3 Foley Video To Audio: the video loader read the frame limit (169) as "skip first frames". Fixed.
LTX2.5 Text To Audio: the text encoder and prompt enhancer were connected to the wrong inputs, and the enhance prompt option held a model file name. Fixed.
LTX2.5 IC LoRA Pixel Spatial Upscaler 2x (FAST Distilled and Slower HQ): the VAE paths now point to the LTX-2 folder, where our model downloader saves them.
LTX2 and LTX2.3 Audio Lip Synch Image To Video (3 presets): now use our new torchaudio-free Mel-Band RoFormer nodes for vocal separation.
AvatarForever Unified Image or Text + Audio To Long Video: now renders an exact 832x1248 portrait by default, instead of a 768x512 pixel budget that followed the input image's shape.
Installer Improvements
Smarter node updates (Windows, RunPod and Massed Compute): the updater now asks GitHub for each node's default branch and fetches only that branch. When we move a node to a different fork, as with ComfyUI-LTXVideo in this release, existing installs switch over cleanly.
Reviewed fork updates: our forks of rgthree-comfy, ComfyUI-Easy-Use, comfyui_controlnet_aux and RES4LYF are synced with their latest upstream code. We read every new upstream commit for safety first, then re-ran our presets with identical prompts: 85 of 85 passed.
comfy-kitchen minimum version raised to 0.2.37, which the new MiniMax H3 sparse attention and long lip sync need.
22 September 2026 Update V138
The folder that holds how LLM instructions renamed into Automatic_Prompt_Generate_Enhance_To_Video_And_Image_Models_LLM_Instructions
New instruction added for Qwen_Image_2.1_Enchance_Prompt_Feed_For_LLMs.txt that supports all model capabilities like transparent background, image references, image editing, inpainting, etc.
New preset Qwen Image 2.1 - Unified Generate Edit Inpaint RGBA Ultimate Quality - Slow - 260922.json added into presets folder
This preset generates highest quality Qwen Image 2.1 images
20 September 2026 Update V136
This is a major workflow and compatibility update. V135 adds three complete unified presets, two new prompt-enhancement guides, a much larger LTX-2.5 prompting guide, a faster LTX-2.5 video decoder, lower-size MiniMax H3 VAE defaults, ComfyUI 0.36.0, and all matching custom-node and SwarmUI extension updates.
Updated Presets Folder Below
NEW PRESET — AvatarForever Unified Image or Text + Audio To Long Video
"AvatarForever Unified Image or Text + Audio To Long Video - 260920"
One workflow now covers image + audio → long avatar video and text + audio → long avatar video. It uses the released AvatarForever audio-conditioned autoregressive pipeline through our SECoursesAudioTools integration.
Load a portrait and any speech or singing recording, describe the visible performance and press Run.
duration_seconds = 0 uses the whole recording. You can trim from any audio start point or select 10, 30, 60, 300 seconds or another duration.
The source soundtrack is preserved and muxed into the final H.264 MP4. This workflow does not generate, replace or transcribe the voice.
Turn Use Image off for audio + text generation. With an image, Match Image Aspect preserves the portrait framing at the chosen pixel budget.
Long videos are generated chunk by chunk. Tiled VAE decoding and streamed export prevent the complete RGB movie from accumulating in memory.
Optional ForeverCache is available, but it can alter output and did not improve end-to-end speed in our RTX 5090 tests, so it remains off by default.
Optional mouth enhancement is enabled by default:
CodeFormer restores aligned face crops, but only a feathered mouth region is pasted back.
Default fidelity is 0.9 and mouth blend is 0.7.
GPU SCRFD face detection and batched FP32 CodeFormer processing are used. Frames without a detected face pass through unchanged.
Turn Mouth Enhancement off to keep the original decoded frames.
A new AvatarForever_Enchance_Prompt_Feed_For_LLMs.txt is included. Give it to ChatGPT, Claude or Gemini together with your portrait and request, and it writes an AvatarForever-compatible visual performance prompt without inventing a transcript.
Use our model downloader app and download AvatarForever Core Bundle (https://www.patreon.com/posts/114517862)
NEW PRESET — Qwen Image 2.1 Unified Generate, Edit, Inpaint and RGBA
"Qwen Image 2.1 - Unified Generate Edit Inpaint RGBA - 260920"
This is one clean Qwen Image 2.1 workflow for text-to-image, multi-image editing, image-to-image, native mask inpainting and transparent-background RGBA generation.
Generate: write a prompt and Run.
Edit: add images in the gallery and address them as @image1, @image2 and so on. Drag to reorder and click X to remove.
Up to 10 images are supported, including the optional init image.
Inpaint: upload an init image, click the native ComfyUI Paint mask button, paint, save, choose Inpaint Masked Area and Run.
The native mask editor supplies brush, eraser, opacity, hardness, invert and undo. Pixels outside the mask are restored from the resized source after generation.
RGBA: enable transparent output and describe a cutout. PNG saving and the gallery preserve the alpha channel.
Text-only generation uses the synchronized canvas controls. Edit modes follow the init or first-reference aspect ratio.
Reference resolution 1024 is the practical default, 2048 uses native 2K references and 0 keeps the original size rounded to 32.
Our preset has native inpainting mask generation
It is also all in 1 preset and 1024px generation takes 3.8 second on RTX 5090 see below image features
Use our model downloader app and download Qwen Image 2.1 Core Bundle (https://www.patreon.com/posts/114517862)
NEW PRESET — Stable Video Infinity 2.0 Pro, Unified Long Video for Wan 2.2
"Stable Video Infinity 2 Pro - Unified Long Video - Wan 22 - 260919"
This new preset generates connected long-form Wan 2.2 video clips while carrying clean motion latents between clips instead of restarting every segment.
Upload a start image and put one clip prompt in each paragraph. A blank line starts the next clip prompt; the last paragraph repeats when more clips are requested.
The default three-clip run creates 233 frames — 14.56 seconds at 16 FPS.
Every clip receives a different deterministic seed derived from the base seed.
Quality mode: 30 steps, CFG 4, Euler/simple, matching Wan 2.2 high/low I2V models and the official SVI Pro adapter pair.
Fast mode: 4 steps, CFG 1 and the selected Seko acceleration pair. It is faster, but can change motion, identity and quality.
motion_latent_count = 1 is the Pro carry recipe. Set it to 0 for independent clips sharing only the anchor.
Continue a source video: enable the muted video loader, choose a video and turn Continue Video on. The original can optionally be included in the final output.
Exact native-latent resume: save the last latent, reconnect it on the next run, keep the anchor and resolution and advance the seed.
Only one decoded clip is held in memory and the MP4 is written incrementally.
SVI Pro produces silent video. SVI-Talk, SVI-Dance and EverAnimate are separate models, not hidden toggles.
A new Stable_Video_Infinity_2_Pro_Enchance_Prompt_Feed_For_LLMs.txt is included. It teaches an LLM the paragraph-per-clip format, real retained-frame timing, continuity planning, source-video extension and honest capability limits.
Use our model downloader app and download Wan 2.2 Core Bundle (https://www.patreon.com/posts/114517862)
UPDATED — LTX-2.5 Audio To Video Is Faster During Decode + Save
"LTX2.5 Audio To Video - Optional Image - 2-Stage 8+3 Steps - 260913"
The sampling recipe, identity anchors, audio timing and two-stage 8+3 generation remain unchanged. The old core VAEDecodeTiled node is replaced with our new SE LTX-2.5 Fast Tiled Video Decode node.
It uses the same tile positions, overlap and feathering as ComfyUI's core decoder.
Decoded tiles are blended on the GPU and the finished video is copied to system RAM once, instead of copying and blending every tile on the CPU.
A 10-second 896×1344 clip saved about 4 seconds in decode + save time.
The default float16 accumulation matched the core result to less than one 8-bit output level in the controlled comparison.
Float32 accumulation is available, but doubles the GPU blend buffer.
For very long 1080p clips on low-VRAM cards, replace it with the core Tiled VAE Decode using the same tile values if the GPU blend buffer is too large.
UPDATED — All 10 MiniMax H3 Presets Now Use the INT8 ConvRot Video VAE
The following presets now select minimax_h3_video_vae_int8_convrot.safetensors instead of the FP16 video VAE:
MiniMax H3 Audio Only With Optional References 4x Speed
MiniMax H3 Image or Video To Video 4x Speed
MiniMax H3 Image or Video To Video 8 Steps
MiniMax H3 Lip Synch — Audio + Text, Image or Video
MiniMax H3 References To Video 4x Speed
MiniMax H3 References To Video 8 Steps
MiniMax H3 Streaming Lip Synch — TaoMate 3 Step
MiniMax H3 Text To Video 4x Speed
MiniMax H3 Text To Video 8 Steps
MiniMax H3 Text With References To Image 4x Speed
The selected video VAE file is approximately 2.95 GB instead of 4.85 GB. The audio VAE, prompts, transformer selections, LoRAs, samplers, schedules and all workflow controls are unchanged by this preset migration.
PROMPT GUIDE — LTX-2.5 Feed Extensively Rewritten
LTX2.5_Enchance_Prompt_Feed_For_LLMs.txt received a major rewrite rather than a small wording update.
Exact map of all ten SECourses LTX-2.5 presets and the prompt method each one needs.
Complete Audio To Video rules for optional images, transcripts, identity anchors, source audio, timing and static-camera identity protection.
Correct CFG behavior: negative prompting is inactive at CFG 1 and active only in workflows sampling above CFG 1.
Correct handling of cached conditioning, local versus hosted encoding and when prompt text leaves the machine.
Stronger person, camera, sequence, dialogue, singing, audio and exact-literal instructions aligned with current LTX-2.5 behavior.
Clear separation between prompt language and runtime controls such as steps, sigmas, keyframes, guides, decoding and upscaling.
HOW TO UPDATE
Download the latest ComfyUI_V135.zip from this post.
Extract it over your existing installer folder and overwrite the older files.
Run Windows_Install_Or_Update_ComfyUI.bat.
Run Windows_Custom_Nodes_Bundles_Installer.bat and install/update bundle 100.
Restart ComfyUI after the updates.
Use our latest SwarmUI Model Downloader and download/update only the bundles you need:
LTX 2.3 Avatar Forever Core Bundle for AvatarForever.
Qwen Image 2.1 Core Bundle for the unified Qwen Image 2.1 preset.
Wan 2.2 Core 4 Steps Bundle for the SVI 2.0 Pro adapter pair and existing Wan models.
MiniMax H3 Core Bundle to obtain the INT8 ConvRot video VAE selected by all updated H3 presets.
SwarmUI Installer and Unified Massive Models Downloader: https://www.patreon.com/posts/114517862
15 September 2026 Update V134
Presets Folder Content In Latest Zip File
NEW PRESET — MiniMax H3 Lip Synch (Audio + Text, Image, or Video)
"MiniMax H3 Lip Synch - Audio + Text Image or Video - 260913"
Drop in an audio file and get a lip-synced video. The soundtrack decides the duration — nothing else needs a frame count.
Three ways to use it:
Audio + text only — H3 invents the performer and the scene
Audio + image — your presenter photo keeps the identity (recommended when identity matters)
Audio + video reference — a new performance guided by that video
What is new inside it:
Automatic H3 prompt formatting. You write a normal scene description; the preset builds the full structured H3 prompt for you.
An Optional Transcript / Lyrics field for the exact words already in your audio. Leave it blank to just follow the audio.
A single Quality / Turbo node with four editable Turbo LoRA selectors: FL2VA 4-step, FL2VA 8-step, Ref2VA 4-step, Ref2VA 8-step. Turning Turbo on routes to the right slot automatically and applies the correct shifts (FL2VA 6/3, Ref2VA 8-step 12/3 — a shift mistake that silently ruins Ref2VA output in hand-built graphs).
A new enable_sla_attention toggle with a keep-percentage and its own SLA LoRA slot (details in the speed section below).
Everything you already had is still there: gallery, references, prompt chains, folder batches, matching batch media, continuation, merging, the extra LoRA stack, preview, audio fallback and the optional face refinement pass.
Measured on the same 10.53 second audio + text clip:
50 steps, no Turbo — 1132.68 s
Same prompt and seed, FL2VA Turbo 4 steps — 124.93 s (9.07x faster)
More measured runs (1024x1024, 24 fps, all successful):
Audio + image, FL2VA Turbo 4, 5 s — 54.98 s
Audio + image, FL2VA Turbo 8, 5 s — 77.14 s
Audio + video reference, Ref2VA Turbo 4, 5 s — 70.82 s
Audio + video reference, Ref2VA Turbo 8, 5 s — 111.38 s
Honest note: with a supplied image, 4-step Turbo kept a clear face, stable identity, background and hand detail in the inspected frames. Text-only Turbo changed the synthesized person's appearance versus the 50-step result. Use an image when identity matters.
NEW PRESET — MiniMax H3 Streaming Lip Synch (any length, TaoMate 3-step)
"MiniMax H3 Streaming Lip Synch - TaoMate 3 Step - Audio + Image Any Length - 260914"
This is the big one. One image + a soundtrack of ANY length = ONE continuous lip-synced video. A 5 minute narration becomes a 5 minute video. No stitching, no per-clip identity resets, and the MP4 is written while it generates.
This is our port of Alibaba TaoLive's TaoMate-H3 streaming runtime onto ComfyUI's native MiniMax H3 blocks. No duplicated model implementation — it drives the real H3 blocks directly.
How it works:
The sampler generates 39 / 34 / 34 / 17-frame chunks (TaoMate's 2, 2, 2, 1 groups per 5 seconds) with a losslessly converted full-rank TaoMate 3-step LoRA.
It keeps a clean KV history — the first chunk as a "sink" plus the two most recent chunks — so every chunk genuinely continues the previous ones instead of restarting.
Your first-frame image stays visible to every chunk, which is what keeps identity over minutes.
Finished chunks are VAE-decoded incrementally and piped to the MP4 as they complete; your original audio is muxed at the end and the file is cut to the exact audio length.
Optional node: H3 Streaming Prompt Schedule — one full H3 prompt per 5 second window, so you can change what the presenter is doing minute by minute, exactly like TaoMate does with one prompt per request.
Measured timing matrix at 512x768 (fp8 KV cache in pinned RAM, resident weights):
1 minute — 291 s
2 minutes — 581 s
3 minutes — 867 s
4 minutes — 1161 s
5 minutes — 1451 s
That is a flat 4.8x real time at every length — it does not degrade as the video gets longer. At 768x1152 a 1 minute run took 694 s (11.5x real time). Per-window transcript prompts cost essentially nothing (291 s vs 288 s for a single generic prompt, plus about 1.2 s of text encoding per window).
Quality check on the 5 minute 512x768 output: identity, framing and lighting hold from 0.5 s to 299.5 s. Fine background text softens slightly in the later minutes. The 2 minute run showed no seam at the decode boundary.
Important settings baked into the defaults:
steps 3 with shifts 12 / 3 — the distilled schedule the LoRA was trained for.
weights = resident. ComfyUI's dynamic weight paging evicts weights as the cache grows; the first 2 minute attempt with paging ran at 36-43 s per chunk instead of 3-4 s.
kv_cache_device = auto — puts the fp8 cache in pinned system RAM when GPU budget is short, which is what makes long and high-resolution runs possible at all.
REBUILT — LTX-2.5 Audio To Video, with a real identity fix
This is a special custom build workflow with custom nodes that we develop which has real idendity fix and it is many times better than public online presets.
"LTX2.5 Audio To Video - Optional Image - 2-Stage 8+3 Steps - 260913" (replaces the 260821 preset)
The sampling recipe is unchanged (official Lightricks two-stage distilled graph: 8 steps at half resolution, 2x latent upscale, 3 refine steps, frozen source-audio tokens). Everything AROUND it was rebuilt.
Graph and usability:
Flat graph with 6 numbered groups instead of 5 nested subgraphs. No API-key text encode path, no Gemma enhancer.
The image is now genuinely optional — the image selector has a real "none" entry. Pick an image = image + audio to video. Leave it on "none" = audio + text to video.
You now type the FINAL output size. Stage 1 is computed for you (half resolution on the 32 px grid), then 2x upscale and an exact center crop to your target.
Pick or upload an image and the target size is derived from its aspect ratio at the 1080p pixel budget automatically: 16:9 → 1920x1080, 9:16 → 1080x1920, 1:1 → 1408x1408, 2:3 → 1152x1728, 4:3 → 1600x1216. Still fully editable, and 0/0 means auto on the backend too.
Whole audio by default: start 0 / duration 0 = the entire file. The frame count is the smallest valid count that covers the audio, and the audio is padded with silence to the exact video length — no more clips that are a few frames short of their own soundtrack.
New warning when the input image aspect and the target aspect do not match (this used to look like a mystery "zoom" on frame 0 — it was just a crop).
lead_in_silence_seconds (default 0.25 s) so the first frame is not mid-word.
Optional SE Torch Compile node between the loader and the guiders, OFF by default and a pure pass-through when off. Turn it on only for long repeated batches.
THE IDENTITY FIX
The old default prompt asked for a "slow, steady push-in" — and that camera move was the single biggest cause of the input face drifting into a different person. We measured 22 variants with ArcFace cosine similarity against the input face, every 8th frame:
Old defaults, 8 s singer clip: 0.694 mean / 0.457 worst — a different person by 6 seconds
Static, locked-off camera prompt alone: 0.718 / 0.586 — zoom gone, identity steadier, zero lip-sync cost
Locked camera + identity anchors: 0.832 / 0.681 — same person the whole way through
What shipped:
A new default prompt: locked-off camera, speaking presenter, "same person throughout".
A new SE LTX-2.5 Identity Anchors node: it re-injects your input image as a keyframe guide every ~4 seconds at strength 0.4, with an attention mask that EXCLUDES the mouth and jaw so lip sync keeps working through the anchor, snapped to the quietest 8-frame block nearby, never inside the last second. Plus the required guide crop after stage 1.
Validation with the shipped defaults:
Presenter photo, 1280x1920 narration — 0.845 mean / 0.723 worst / 0.87 at the end
Singer avatar 8 s — 0.777 / 0.684 (was 0.694 / 0.457)
Singer avatar 18 s — 0.737 / 0.510 (was 0.680 / 0.534), and now the same person for all 18 seconds instead of changing face at around 5 s
Audio only, no image — untouched pass-through
Things we tested and REJECTED so you do not waste your time: first-frame strength 1.0, plain euler, lower stage-2 sigmas, doubling stage-1 pixels, and switching stage 1 to the Dev transformer (2.3x slower, four runs spanned 0.62 to 0.79, and it follows camera prompts so literally the clip ends in an extreme mouth close-up).
Measured batch speeds per 7-8 second segment with an input image (2:3 outputs):
1280x1920 — 94 s
1024x1536 — 62 s
896x1344 — 42 s
768x1152 — about 32-35 s
Tip for batching: keep clips at or below 10 seconds. Compute grows faster than linearly with duration, and 15 s at 1280x1920 already runs out of VRAM on 32 GB so it may use shared VRAM and gets too slow.
CUSTOM NODE PACKS WE DEVELOP UPDATED
SE LTX-2.5 Input Image (optional) + Target Resolution
SE LTX-2.5 Audio → Frozen Latent + Frames
SE LTX-2.5 Image Conditioning (auto skip)
SE LTX-2.5 Identity Anchors (keep the same face)
SE LTX-2.5 Save / Load Conditioning (.pt cache) — skip re-encoding the same prompt across runs
SE Torch Compile (optional, off by default)
SE Fit Frames To Exact Size (center crop)
H3 Lip Synch - Automatic Prompt
H3 Lip Synch - Quality / Turbo
H3 Streaming Sampler - TaoMate 3-Step (Audio + Image → Long Video)
H3 Streaming Prompt Schedule (one prompt per 5 s window)
Plus a batching optimization: the insightface detector is now created once per process and the mouth mask is cached per input image, so a batch that reuses a few presenter photos no longer pays for face detection on every single job.
ComfyUI-TeaCache (our fork) — three upgrades, all for MiniMax H3:
SOL sparse attention rebuilt on native Comfy Kitchen SOL. On real H3 tensors (37,741 tokens, 56 heads) the new path is 2.1x to 2.6x faster than our bundled kernel AND 22-37% more accurate in the sampled calls. The auto selector timed Kitchen at 43.30 ms vs SageAttention 81.86 ms vs bundled SOL 110.40 ms and picked Kitchen. 256 extra routed tokens are now enabled, which reduces brightness and detail pulsing.
New sparse_dense_last_steps control (default 1) keeps the final detail step dense. FirstBlockCache now always computes the final step, including on 4 and 8 step schedules — previously a fractional end bound did not guarantee this.
Two real bug fixes: FirstBlockCache tensors are now allocated outside ComfyUI's new AIMDO allocation compiler scope (zero rogue allocations, no AIMDO error), and enabling SOL together with the low-VRAM save path no longer throws a TypeError — the low-VRAM block now forwards ComfyUI's attention override like the stock block does.
PROMPT FEED v3.1 (MiniMax H3)
MiniMax_H3_Enchance_Prompt_Feed_For_LLMs.txt updated and grew by roughly 470 lines. Paste it into your LLM and it writes correct H3 prompts for you. New in this edition:
A full new chapter on directing a PERFORMANCE: a cause-and-visible-change method (starting state → trigger → eye and face response → body and voice response → settle) with a table of concrete emotional turns and the audible behaviour that matches each one.
A bounded inventory of 25 community dialogue cues (<pause>, <breath>, <inhale>, <i>word</i>, <whisper>, <sighs>, <laughs>, <gasp>, <stutter>, <softer> and more) with strict rules for using them sparingly — and an explicit ban on ever adding them to an exact script or a locked soundtrack.
Correct handling of the locked-soundtrack path: prompt tags CANNOT insert a pause or a sigh into audio you supplied, so the guide now tells the LLM to choreograph the face around your actual recording instead.
Rules for the multi-frame continuation context (1, 5, 22, 39 or 56 previous frames), including budgeting words around the overlap that gets removed on merge.
Reference label normalization notes and a hard rule against inventing H3 fields like "emotion:" or <sadness=0.8>.
HOW TO UPDATE
Download the latest zip from this post and extract it over your existing installation folder.
Extract and overwrite all files
Run Windows_Install_Or_Update_ComfyUI.bat file
Then run Windows_Custom_Nodes_Bundles_Installer.bat install bundle 100
Also to get new few LoRAs use SwarmUI model downloader app, get LTX 2.5 and MiniMax H3 core bundles, it will only download missing new files : https://www.patreon.com/SECourses/posts/swarmui-auto-and-114517862
27 August 2026 Update V132
Added the compact + Prompt inline prompt chain to the Reference Gallery.
Multiple prompts can now run sequentially while sharing references, continuation settings, and optional merged output.
Reference tokens are now case-insensitive and tolerate common aliases and spacing variations.
E.g. <Picture 1>, <picture 1>, @image1 all works
Native MiniMax labels such as <picture1> are normalized to the exact format expected by the model.
Get latest zip file, overwrite all and run installer and bundle installer bat file to update
26 August 2026 Update V131
When you touch the prompt box, it was becoming shrinked and this issue resolved
MiniMax H3 Upgrades
Continue videos using the final 1, 5, 22, 39, or 56 frames, instead of only the last frame.
Multi-frame continuation uses MiniMax H3’s native clip guide for stronger motion, identity, and scene continuity.
Replayed context frames and matching audio are automatically removed from the generated segment before merging.
Continuation works with both FL2VA and Ref2VA without consuming an additional reference-image slot.
The two Image To Video presets are now Image or Video To Video.
So with these presets now you can do image to video or video to video
Start Frame / Init Video now accepts images or videos. A video can supply its final frame or selected multi-frame context.
Merge videos can append the new generation directly to the uploaded source video, preserving video order and audio.
Folder batches retain sequential generation, per-prompt saves, continuation, and final merged outputs.
Now we have contextual continue feature with Continue From Last Video Frames option we have
This feature implemented into SwarmUI as well so read it too
To update : get latest zip file, extract and overwrite all, then run Windows_Install_Or_Update_ComfyUI.bat and also Windows_Custom_Nodes_Bundles_Installer.bat (bundle 100) - all necessary custom nodes and extensions automatically installed and many of them managed by us for extra security
22 August 2026 Update V129
Python 3.10 not anymore supported because onnxruntime-gpu 1.27+ CUDA 13 build requires Python 3.11 or newer
The installer will auto delete your existing Python 3.10 venv and auto generate new Python 3.12 venv if you have Python 3.12
Please have Python 3.12 and use it
This is a very big update : 8 new official LTX 2.5 presets, MiniMax H3 Init Audio (make the video follow any soundtrack with lipsync), live token meter, Face Inpainting upgrades, automatic FL2VA / Ref2VA routing, a speed update and a big installer overhaul so please read all
8 new LTX 2.5 presets added
The 2-stage presets generate at 960x544 with 8 distilled steps, then apply the official LTX 2.5 latent spatial upscaler x2 and run 3 more refinement steps on a 1920x1088 canvas, and finally center crop to exact 1920x1080 - all with synchronized audio at 24 FPS and about 5 seconds by default
The Fast preset stays single stage at 960x544 for quick tests, and the First Last Frame preset is the official single-stage 1280x720 workflow : load a first and a last image and it generates the connected shots in between
The Expert preset exposes every stage on the canvas (Preprocess, Generate stage 1, Upscale and re-sampler stage 2, tiled Decode) with notes explaining each setting - the Text To Video and Image To Video presets are the compact subgraph versions, enter the subgraph when you want to fine tune
Audio To Video : load any audio, trim it with audio start and duration, and the video is generated to follow it (lipsync, action timing) - the encoded audio tokens stay frozen in both stages and your original trimmed audio is muxed into the output untouched, never decoded from latents - an optional start image is supported
Motion Track Control : drag points on the LTX Sparse Track Editor to sketch how things should move on your start image, and the motion track IC-LoRA animates exactly along those paths - a tracks preview video is saved next to the result so you can check the motion before committing to a long run
Text To Audio : audio only generation (sound effects, ambience, music cues) with the same LTX 2.5 model - no video VAE and no upscaler needed in this flow
Prompt enhancement (the official Gemma 4 E2B INT8 ConvRot enhancer) is off by default in every preset - enable it to expand short prompts into rich cinematic ones - and entering an LTX API key switches text encoding to the API to save VRAM
Models : one bundle for everything - open the latest model downloader and grab the LTX 2.5 Video Core Bundle : it now contains every model referenced by all ten of these LTX 2.5 presets : https://www.patreon.com/posts/114517862
LTX2.5 IC LoRA Pixel Spatial Upscaler presets fixed (FAST Distilled and Slower HQ) : the 4 helper nodes they need (LTX-2.5 Upscale Controls, Distilled Sigma Schedule, Prepare Video Chunks, Merge Video Chunks) are now part of our auto installed and managed by us extensions
Both upscaler presets also switched from the 24.46 GB BF16 text encoder to the 14.32 GB Gemma 4 12B INT8 ConvRot v2 one, so all ten LTX 2.5 presets now share the exact same text encoder and prompt enhancer - that is 10 GB less to download and one less model kept in memory when you switch between presets
New LTX2.5_Enchance_Prompt_Feed_For_LLMs.txt inside the Prompt_Generate_MiniMax_LTX_and_Demo_Materials folder : an official-aligned LTX 2.5 prompting guide for LLMs - give it to ChatGPT / Claude / Gemini, tell it what you want, and it writes properly formatted LTX 2.5 prompts for text / image / audio to video, multi-shot and IC-LoRA workflows
Optional Init Audio added to all six MiniMax H3 video presets (Text To Video, Image To Video, References To Video - both 8 Steps and 4x Speed) : select or upload any audio - or even a video file - in the new INIT AUDIO (OPTIONAL) node and the generated video follows that soundtrack exactly : lipsync, action timing, ambience - and keeps it as the exact final audio track
Leave it at (none - disabled) and the presets work exactly as before
Works with text only prompts, start frame and gallery references - by default the video duration automatically follows the audio length or set duration_mode to keep workflow duration
Init audio accepts video files with a visual trim panel - drag the handles to pick exactly which part of the soundtrack to use, nothing is re-encoded
Folder batches : put a same-name audio / video file next to each batch item and it is automatically used as that item's init audio (SAME-NAME INIT MEDIA)
Live MiniMax H3 token meter added to the reference gallery : prompt + references + resolution + duration are counted in real time against the 109k token budget so you see the usage before you queue - the meter also follows your init audio selection and trim window
Automatic FL2VA / Ref2VA routing : the MiniMax H3 video presets now pick the right checkpoint per prompt by themselves - single runs that have references and every folder batch item go through the Auto FL2VA / Ref2VA pipeline while plain text prompts keep the normal path - and the References To Video presets got an optional INIT IMAGE node : FL2VA uses it as the exact first frame, Ref2VA adds it as the starting frame picture reference (folder batch items ignore it and keep using their same-name init media and last frame continuation)
Face Inpainting upgrades : new faces to inpaint widget selects which faces get refined : 1 = biggest face (default), 2 = second biggest, 1,3 or all - faces are ranked biggest to smallest and each selected face is refined in its own pass and tagged in the tracking preview
New hallucination guard : every regenerated face crop is compared against its input and frames where H3 painted a neighbouring or different face over your subject are not pasted
Geometry lock is further improved and ON by default - removes the per-frame face shaking / tilting that the face pass could introduce
MiniMax H3 Text With References To Image preset also received the new face inpaint subgraph and a randomized seed
Speed : MiniMax H3 Sol-Attn sparse attention updated for SM89 (RTX 4000 series) and CUTLASS 4.7 in our maintained ComfyUI-TeaCache fork
Example on RTX 5090 : 1344x768 10 second video dropped from 450 seconds to 401 seconds with the updated speed optimizer
FLUX 2 Klein presets switched to my new INT8 ConvRot HQ models - almost BF16 quality but massively faster - auto download them with the latest model downloader : https://www.patreon.com/posts/114517862
Installer overhaul : custom nodes are now updated in-place instead of delete and re-clone so files downloaded inside node folders (for example ckpts) survive updates - a fresh clone replaces a node only when the in-place update fails
Leftovers of interrupted installer runs are cleaned automatically and staged copies use .disabled names so ComfyUI never imports duplicate nodes
Deleted venv is recreated in place without re-cloning ComfyUI - your models, outputs and user data are never touched - and the installer window now always stays open with a clear final status message
Python 3.10 support dropped because onnxruntime-gpu 1.27+ CUDA 13 build requires Python 3.11 or newer : menu now offers 3.11 / 3.12 (recommended) / 3.13
OpenCV conflicts between custom nodes fixed for good : all four OpenCV distributions are kept at one resolved version and opencv-contrib-python-headless is reinstalled last then verified - RES4LYF, controlnet_aux, Frame-Interpolation and rembg each pull different ones
onnxruntime-gpu upgraded to 1.27+ (CUDA runtime used by FantasyPortrait) and google-genai added to requirements
ComfyUI-Frame-Interpolation now installs without CuPy - it is not needed by the bundled RIFE workflows
RunPod and Massed Compute installers received the same in-place update and OpenCV normalization treatment
Windows_Run_GPU.bat renamed to Windows_Run_ComfyUI_Default.bat
To update : get latest zip file, extract and overwrite all, then run Windows_Install_Or_Update_ComfyUI.bat and also Windows_Custom_Nodes_Bundles_Installer.bat (bundle 100) - all necessary custom nodes and extensions automatically installed and many of them managed by us for extra security
16 August 2026 Update V122
Presets and custom nodes updated to improve face shaking effect that happens when automatic face inpainting activated
All presets now do have live preview (no increase in VRAM or generation duration) and fixed seed feature
16 August 2026 Update V120
MiniMax H3 References To Video 4x Speed - 24 FPS - 260808.json was broken with mistake and now fixed
Newest speed LoRA for MiniMax H3 is now in presets from LightX2V
Use latest Model downloader to get updated bundle : https://www.patreon.com/SECourses/posts/swarmui-auto-and-114517862
I have added automatic Face Inpainting feature to all MiniMax H3 presets
Get latest zip file, extract and overwrite all and run installer update bat file to update
Also use updated extra_model_paths.yaml if you are downloading models into SwarmUI
To understand how Face Inpainting works see this video clip
Live preview node added to all presets and does not cause any slowness or extra VRAM usage
14 August 2026 Update V117
Since you have requested, all MiniMax H3 presets modified and easy LoRA use added
If you don't select any LoRA, workflow works just fine no issues so they don't break presets and how they work, they are optional
3 New presets added
MiniMax Music 3 Text To Music - 260813.json - New model published today
LTX2.5 IC LoRA Pixel Spatial Upscaler 2x - FAST Distilled 8 Steps - 260813.json
LTX2.5 IC LoRA Pixel Spatial Upscaler 2x - Slower HQ - 260812.json
MiniMax Music 3 is reportedly lower quality than ACESTEP XL 1.5
Our very advanced ACESTEP XL 1.5 Premium app below and it has full LoRA and DoRA training support
MiniMax Music 3 also missing necessary resources to be trainable
MiniMax Music 3 prompt templates shared here : https://github.com/MiniMax-AI/MiniMax-Music3/tree/main/skills/music-caption-rewriter/templates
LTX 2.5 video upscaler preset is able to upscale any video, any resolution, any aspect ratio into 2x
It is a very serious generative upscale so adds lots of details
Therefore, try to upscale at 1 chunk to have better consistency
Read preset information carefully after loading into ComfyUI
Our SECourses Premium Upscaler Pro app also fully supports LTX 2.5 now : https://www.patreon.com/SECourses/posts/secourses-pro-150202809
It is standalone does not require ComfyUI and it auto downloads models when you first time use
To auto download necessary models of these presets, use our model downloader app and download from ComfyUI bundles : https://www.patreon.com/SECourses/posts/swarmui-auto-and-114517862
Moreover, I have converted the LTX 2.5 Int8 ConvRot myself higher quality and more accurate than officially published ones
Someone detected one of the model is corrupt and broken - officially released ConvRot Int8 variant
I converted them from BF16 versions after big research and experimentation
So our model downloader app will download better quality Int8 ConvRot of LTX 2.5 models
Our Musubi Trainer convert to quant tab also now supporting LTX 2.5 Int8 ConvRot compile : https://www.patreon.com/SECourses/posts/secourses-musubi-137551634
LTX 2.5 video generation presets and bundles will be added soon hopefully but its quality lower than MiniMax H3
Get latest zip file, extract and overwrite all, then run Windows_Install_Or_Update_ComfyUI.bat and Windows_Custom_Nodes_Bundles_Installer.bat (select bundle 100) for updating all
12 August 2026 Update v116
To improve security, I have forked auto installed repos and bundle 100 repos and now they will be installed from secure fork
Only few repos left as exception which are famous : mcmonkeyprojects , kijai , Lightricks
I also did a massive deep security check of all auto installed and bundle 100 repos source codes and all are secure
MiniMax H3 presets got updated and now we have new features
Now when batch processing, you can dynamically set the duration of each prompt
It will split the file name of the prompts with _ and if last split is a valid integer, it will be used as the duration of the prompt, otherwise whatever you set will be used
Folder durations now use the final positive integer suffix only when the filename contains : scene7.txt → 7s; 200.txt, scene.txt, and scene_0.txt use the configured duration
For example scene1.txt or 1.txt will use your set duration on the workflow meanwhile scene_1.txt will be set to 1 second duration dynamically
Added Continue from last frame beside Merge, disabled by default
Continuation automatically uses Ref2VA when references exist and FL2VA otherwise
Updated all MiniMax presets, extension UI, documentation, and the LLM prompt feed
MiniMax_H3_Enchance_Prompt_Feed_For_LLMs.txt improved big time and includes latest changes and features as well
The most professional speed up LoRA maker LightX2V just published new speed LoRAs for the MiniMax H3
Our MiniMax H3 core bundles now auto downloads both FL2VA Turbo 8-step v1.0 and FL2VA Turbo 4-step v1.0 768p
Use model downloader to auto download : https://www.patreon.com/SECourses/posts/swarmui-auto-and-114517862
All presets by default uses FL2VA Turbo 8-step v1.0 but you can switch it to FL2VA Turbo 4-step v1.0 768p
Older version 0.1 removed from downloads and presets so you can delete from LoRA folder if you downloaded
All presets exposes Shift value for Video and Audio seperately now
So you can set to 6 / 3 if you switch to FL2VA Turbo 4-step v1.0 768p - default is 12 / 3
Sadly LoRAs only made for non-reference model atm but works on both models until they publish specific new LoRA for reference model
Some people asked me if video-latent continuation input is possible or not and sadly not possible so your best option is using continue from last frame
Get latest zip file, extract and overwrite all, then run Windows_Install_Or_Update_ComfyUI.bat and Windows_Custom_Nodes_Bundles_Installer.bat (select bundle 100) for updating all
11 August 2026 Update v113
RunPod and Massed Compute startup command changed so that the running ComfyUI instance won't be accessible by public IP anymore
Updated files are : Massed_Compute_Instructions_READ.txt and RunPod_SimplePod_ComfyUI_Instructions_READ.txt
Windows doesn't have such vulnerability it is fully locally running
Use Cloudflared secure HTTPs to access publicly when you are running on cloud services
RunPod and Massed Compute installers auto installs Cloudflare
Follow Massed_Compute_Instructions_READ.txt and RunPod_SimplePod_ComfyUI_Instructions_READ.txt
Public access is optional, it will be default fully locally restricted running
When using unicode character containing references it was causing issues
For example an image with Ü Ğ characters, this issue fixed and now works
Just run Windows_Install_Or_Update_ComfyUI.bat to update and always get latest zip file and overwrite older files
10 August 2026 Update v112
MiniMax H3 image to video presets upgraded and now they support ending frame as well
If you don't provide ending frame, it will work as before image to video
So init image is first frame, ending image is last frame
If you want to use images as reference, use reference presets
Windows_Run_GPU_Lowest_VRAM.bat added into zip file
This bat file starts the comfyui with following args which you can use in SwarmUI backend as well :
--cache-none --lowvram --disable-pinned-memory --use-sage-attention --enable-triton-backend --disable-smart-memory --disable-dynamic-vram
Extract latest zip file, run Windows_Install_Or_Update_ComfyUI.bat to update / install
For starting regular, use Windows_Run_GPU.bat
Presets updated with new ltx-2.3-spatial-upscaler-x2-1.1.safetensors instead of ltx-2.3-spatial-upscaler-x2-1.0.safetensors - used updated SwarmUI model downloader to download it
9 August 2026 Update v109
Batch folder processing significantly improved
Now you can upload up to 99 images, videos or audio clips
When referencing, the app will use whatever you have in your prompt referenced like @audio5 it will use the @audio5
Still, up to 3 per generation / prompt can be used for audios and video clips and 9 images
This way, you can generate complete videos
To reflect newest features, MiniMax_H3_Enchance_Prompt_Feed_For_LLMs.txt is updated
You can give this MiniMax_H3_Enchance_Prompt_Feed_For_LLMs.txt file to ChatGPT, tell what you want, and it will generate you properly formatted prompts
The file is inside Prompt_Generate_MiniMax_LTX_and_Demo_Materials folder
Inside Prompt_Generate_MiniMax_LTX_and_Demo_Materials folder you will see Example_Prompts_Folder_For_MiniMax_H3_5_Characters_5_Audio_References which contains 1 to 5.txt prompts for 1 minute batch folder clip generation
Excellent example that shows how MiniMax H3 References To Video presets can be used to generate long consistent audio and character long videos
9 August 2026 Update v108
Model downloader app updated and Int4 ConvRot MiniMax H3 models bundle added as Low VRAM - https://www.patreon.com/SECourses/posts/114517862
This bundle is great for 12 GB and below GPUs
At the presets both ComfyUI and SwarmUI, just replace Int8 Model with below Int4 variants
This format Int4 ConvRot models right out of the box working with our ComfyUI (use our installers) backend and installers for SwarmUI
We are using official code of Lightricks not Kijai implementation therefore our LoRA works better
However, Kijai had a VRAM optimization and i just implemented it
You will see it like below in presets, MAX savings reduces VRAM more than 40% but may reduce quality, exact saves 15%+ but same quality
This feature exists in SwarmUI as well with newest update
You can use MAX settings for reference instead MATCH, its much better likeness of your reference image
So you can set to Max from above screenshot from match to max
The importance of Max option is very significant read below image
But be carefull dont add to large photo files, this can slow down everything
Video clip references are processed differently and below explains how they processed when used as a reference
8 August 2026 Update v107
This is a very major upgrade with so many new amazing stuff so please read carefully
Famous Lightricks released 4 steps LoRA for MiniMax H3
I have fully and accurately implemented it into our installers
Currently not working with SwarmUI or ComfyUI as default but works with our installers due to myself developed premium nodes and extensions
Make sure to install update your ComfyUI and SwarmUI users must also use our ComfyUI backend
This LoRA added into our model downloader MiniMax H3 core bundle please download
I have made new presets that uses this LoRA and does 8 steps so now you can use them
It can go as low as 4 steps but I recommend 8 steps
New updated presets are as below
Now all presets have Sampler and Scheduler exposed so you can change them if you wish
4x Speed presets are using https://nvlabs.github.io/Sana/Sol-Engine/H3-OnDevice/ but they are actually 1.39x - named as 4x since Sana officially claimed 4x speed under certain cases
8 steps literally reduces generation time more than 50%
Add references and write prompt field improved so that now you can add videos and audios with trim as you wish as below (trimming is optional)
For @video1's soundtrack, type <Audio 1>; @audio1 is the first standalone audio file
According to the above rule, MiniMax_H3_Enchance_Prompt_Feed_For_LLMs.txt file improved and literally working amazing
It is located inside Prompt_Generate_MiniMax_LTX_and_Demo_Materials folder inside zip file
All presets now have Folder batch processing feature with merge videos option
Batch processing logic is as below
With batch folder processing you can generate even entire movies
I have generated this music video clip : (link will be added and will be explained in tutorial hopefully)
Final merged video may not appear on ComfyUI interface so check output folder to get it
Still individually generated files will be saved as well
Audio only preset implemented which works great
It is a quality-first 32 kHz stereo FLAC workflow using a disposable 32x32 video canvas, 50 res_multistep/beta steps, with the speed optimizer disabled by default.
It automatically selects FL2VA for text-only prompts and Ref2VA when references exist; video references contribute soundtrack only, not frames.
So if you use a video as a reference for audio generation, only audio of it will be used
Audio generation can be used to see if your prompt and duration matching before generating video since it is like real 2x time speed
So 1 minute audio generation takes like 30 seconds on RTX 5090 therefore use audio generation preset to see if your prompt and duration matches - for speaking having videos
So quickly iterate, see if your audio accurate, then generate full video
Moreover, I recommend you to use 8 steps and 0.4 megapixel resolution to quickly generate your videos, verify they are accurate, then move full generation like 1344x768px and 20 steps high quality
This works great
For updating get latest zip file, extract, overwrite all, run Windows_Install_Or_Update_ComfyUI.bat and then run Windows_Custom_Nodes_Bundles_Installer.bat and select bundle 100
6 August 2026 Update v105
We have significantly improved our multi mixed reference uploader custom node - auto installed with our installer
Now when you add or remove an attachment, it will not modify and alter your written prompt
Now you can drag and drop attachments to change their position like @image1 into @image3 etc
Now even if you have reference in the prompt but not in attachment, it will not throw error and at the back end it will remove @image1 etc and prompt will just run fine
Get latest zip file, extract all, run Windows_Install_Or_Update_ComfyUI.bat to update
Also I published a speed up quality comparison video : https://www.reddit.com/r/comfyui/comments/1vh5wd8/i_have_implemented_solattn_crossstep_cache_from/
6 August 2026 Update v104
I have fully and accurately implemented https://nvlabs.github.io/Sana/Sol-Engine/H3-OnDevice/ into our ComfyUI installation via our always installed and myself developed custom nodes
Deploy MiniMax-H3 with Sol Engine achieving up to 4.52× acceleration on on-device GPUs
Their claimed 4.52x speed acceleration happens at 50 steps and default attention meanwhile I measured with 20 steps and against Sage Attention 2.8.3 latest (already speeds up like 20%+)
Moreover, I did set more conservative but almost same quality 0.08 thus it skips lesser steps
The presets are also unified and now we have 4 presets that supports Sol Engine 4x Speed up and multiple mixed references (9x images, 3x videos, 3x audios)
Presets are as below:
MiniMax H3 Text With References To Image 4x Speed : You can generate image with MiniMax 3 with just prompt or with adding reference images to the prompt
MiniMax H3 Image To Video 4x Speed : You can generate videos from initial image with prompt
MiniMax H3 References To Video 4x Speed : You can generate videos with just text or with up to 9x image, 3x audio and 3x video references (optional)
MiniMax H3 Text To Video 4x Speed : You can generate videos with just text prompts
Default values quality wise conservative so almost same quality and brings 1.39x speed compared to Sage Attention 2 speeds at 1344x768 with 362 frames
It used only 20 GB vram for 1344x768 px at 362 frames - 15 seconds but it can use even lower depending on your GPU VRAM
The speed up is not enabled by default in presets so click toggle and enable if you want
The first run will be slow due to kernel compile but consquent runs will be faster and won't recompile
In logs first 4 steps are same, then step speed increases and then you start to see jumps in steps like below
For updating get latest zip file, overwrite all files, and then run Windows_Install_Or_Update_ComfyUI.bat file
5 August 2026 Update v103
We have officially entered enhanced ComfyUI with this update, with my developed and auto installed custom nodes for MiniMax H3
MiniMax H3 added to SwarmUI model downloader and our ComfyUI presets - download core bundle
Can generate up to 30 seconds and 2K native resolution just fine
Windows_Run_GPU.bat now starts with --disable-pinned-memory because pinned memory system is still bugged - bugs reported
Following presets added to presets folder
These presets are extremely significantly enhanced compared to officially released presets by ComfyUI see below images
More variables exposed and you can set any resolution as you wish directly
As you change aspect ratio or megapixels, the width and height are automatically updated and snyched - it uses the width and height at final generation
Providing multiple references was extremely hard but not anymore with our amazing extension
Up to 9 images, 3 audios and 3 video clips fully dynamically supported with all reference taking presets
Our system uses references as @image1 @image2 @video1 @video2 @audio1 @audio2 and so on
Up to 9 images, 3 audios and 3 videos
<audio 1> means the audio of the first input <video 1>
MiniMax_H3_Enchance_Prompt_Feed_For_LLMs.txt file is modified to instruct LLM to write prompts according to our system (as mentioned above) so you will get perfectly ready copy to paste prompts
The txt file is inside Prompt_Generate_MiniMax_LTX_and_Demo_Materials folder
Get latest zip file, extract and overwrite all, run Windows_Install_Or_Update_ComfyUI.bat and install bundle 100 via Windows_Custom_Nodes_Bundles_Installer.bat
1 August 2026 Update
Ultralytics version pinned to 8.3.216 so that even older Yolo models will work
29 July 2026 Update
The zip file is same
I have updated myself maintained fork of ComfyUI-TeaCache and fixed all recent errors that recent ComfyUI changes introduced
Just run Windows_Install_Or_Update_ComfyUI.bat to fix the issues
19 July 2026 V100 Update
This is a pretty significant update so please read carefully and entirely
I have been waiting Torch 2.13 to update our apps and libraries and it recently arrived
So I have pre-compiled the following libraries with Consumer GPUs CUDA archs for Windows and All GPUs CUDA archs for Linux as below
xFormers, TorchAO, Sage Attention, Segment Anything, Segment Anything 2 (SAM2), MSLK (xFormers won't work properly without this and this makes big difference), InsightFace, Flash Attention
None of these libraries do not exists for Windows and Linux wheels are not including all GPU CUDA archs pre-compiled like i do
So now our ComfyUI installs all above libraries with Torch 2.13 and CUDA 13
I recommend to use Python 3.12 during installation but we support all Python 3.10, 3.11, 3.12 and 3.13 since all wheels are now compiled as ABI3
Since we upgrade to new torch, please delete venv and run installer again for update or do a fresh install
Make sure to get latest zip file and overwrite all previous files in installation folder before update
Windows_Run_GPU.bat starts with --use-sage-attention --enable-triton-backend
--use-sage-attention really useful for video models but in some image models may reduce quality so you can test on and off
Our compiled Sage Attention is latest version of 10 July so has lots of fixes
Automatically installed ComfyUI-QuantOps updated
Now it is only used if native ComfyUI is not supporting the loaded model
I have compared the new famous Int8 ConvRot of Krea 2 and the speed difference is like 100%, I plan to update all models to Int8 ConvRot
New updated ComfyUI-QuantOps supports Krea 2 GGUF as well
When you click and see full size of above image and analyze results you will see that:
Int8 ConvRot is 96.2% similar to BF16 meanwhile GGUF Q8 is only 90.0% and FP8 Scaled is 82.2% and NVFP4 is 63.7%
Moreover, Int8 ConvRot generates the output in 3.05 seconds, making it 1.82× faster than BF16, which takes 5.56 seconds.
NVFP4 takes 3.8 seconds and is 1.46× faster than BF16, whereas GGUF Q8 takes 6.06 seconds and is approximately 8.3% slower than BF16.
So Int8 ConvRot generated with our Musubi Trainer app at high quality is almost 100% faster and almost same quality as BF16
High quality generation takes few hours on RTX 5090
Updated SwarmUI downloader has amazing quality Int8 ConvRot which reaches almost BF16 quality : https://www.patreon.com/SECourses/posts/swarmui-auto-and-114517862
Krea 2 Core bundle downloads this model and uses it in SwarmUI preset
ComfyUI preset also uses that model default
You can generate Int8 ConvRot models with our updated Musubi Trainer app : https://www.patreon.com/SECourses/posts/secourses-musubi-137551634
I also have converted LTX 2.3 models into Int8 ConvRot HQ with our Musubi Trainer app (above link)
So Int8 ConvRot HQ is 100% faster than FP8 Quant Scaled and 50% faster than BF16 on RTX 5090
The quality is also excellent almost same as BF16
New Krea 2 Presets - Supported in SwarmUI too with SwarmUI presets
Krea 2 Raw Base - 52 Steps - 260716
Krea 2 Turbo - 8 Steps - 260716
Krea 2 Turbo Image Edit - 8 Steps - 260716
All Krea 2 presets uses compiled by me Int8 ConvRot HQ (each compile takes 3-4 hours on RTX 5090)
New Phantom Wan 14B Character Reference Presets - Supported in SwarmUI too with SwarmUI preset
These presets are for you to upload multiple input images and use them in generated text to video prompts
Up to 6 images supported but recommended are up to 4
Phantom Wan 14B Character Reference T2V Quality - 260717.json, Phantom Wan 14B Character Reference T2V Fast - 260717.json
Please install Bundle 100 via Windows_Custom_Nodes_Bundles_Installer.bat and use SwarmUI model downloader (https://www.patreon.com/posts/114517862) and download 🎬 Phantom Wan 14B Character Reference T2V Bundle (Total: 20.78 GB, 4 models) SwarmUI bundle
New Insanely Powerful LTX2.3 Licon MSR V2 Multi Subject Reference Preset - Supported in SwarmUI too with SwarmUI preset
This preset uses LTX 2.3 and supports 1-4 input image and 1 background image (mandatory to provide atm)
It is literally insanely powerful and generate video fast with LTX 2.3 and very high quality see below very primitive example and how it turned out to be great
New LTX 2.3 Foley Video-to-Audio Workflows
This workflow adds amazing natural audio based on prompt and input video to any silent or audio having video (existing audio is overwritten)
It has intelligent auto chunk system so you can process any duration video
Full Updated ComfyUI Presets As Below click to See Full Size - 63 Local free Presets
All Krea 2 and LTX 2.3 presets now using Int8 ConvRot HQ compiled by me
Each compile takes around 3-4 hours on RTX 5090 but their quality better than GGUF Q8 and they are more than 100% faster than FP8 Scaled on RTX 3000, 4000 and 5000 series (not tested on RTX 2000 and 1000 series yet)
V93 adds synchronized Foley and environmental audio generation from an existing video:
LTX 2.3 Foley Video To Audio - 30 Steps: a standard 24 FPS workflow that automatically trims the input to a valid LTX frame count, keeps the video frozen, generates only the audio modality, and saves the result as an H.264 MP4.
LTX 2.3 Foley Auto Long Video - 24 FPS: designed for longer clips. It streams the source video instead of decoding the full video into memory, processes overlapping 89-frame windows at 576 x 576, crossfades and stitches the generated audio, and muxes it back onto the original-resolution video.
FoleyExtension Is Now Installed Automatically so Foley works perfect
You can see example generated video here : https://www.reddit.com/r/SECourses/comments/1uyk2d8/ltx_23_foley_video_to_audio_video_is_now_fully/
Original used silent video is here : https://www.pexels.com/video/serene-riverside-reflection-in-forest-35591199/
Used prompt : Natural synchronized ambience of a calm river flowing gently through lush greenery, with soft water ripples, a light breeze moving leaves and grass, occasional distant birds and insects, and realistic outdoor spatial depth. No speech. No music.,
I developed custom node for this and working perfect auto installed for you
More Reliable Installers
The Windows and cloud installers received a large reliability pass:
RunPod and Massed Compute now use a stable Python 3.12 virtual environment, reuse it when valid, and only install or upgrade Python when necessary.
The cloud installers now detect root/sudo availability and handle the stable Python 3.12 package source more carefully.
ComfyUI IPAdapter Plus is now installed by default by the cloud scripts.
Custom-node requirements consistently use UV's best-match index strategy.
Protected core requirements are restored after optional custom-node installation so node packages cannot silently leave incompatible ONNX or acceleration libraries behind.
CPU and GPU ONNX Runtime conflicts are cleaned before the protected GPU stack is restored.
The Windows bundle selector now correctly accepts space-separated selections as advertised.
ReActor's Impact Pack dependency now runs its `install.py` when needed.
Paths are anchored to the installer location, making the scripts safer to launch from a different working directory.
Windows now recommends Python 3.12.10 while retaining the 3.10, 3.11, 3.12, and 3.13 choices.
Isolated SwarmUI and Foley Nodes
SwarmComfyCommon, SwarmComfyExtra, and FoleyExtension now live under:
ComfyUI/StandaloneCustomNodes
The updated launchers create and load `ComfyUI/standalone_swarm_nodes.yaml` automatically. Legacy root-level copies and duplicate copies under normal `custom_nodes` are removed only after the new installation has been verified.
This keeps the portable nodes isolated, avoids duplicate `NODE_CLASS_MAPPINGS`, and makes both the normal GPU launcher and the VRAM-optimized launcher use the same configuration.
Before this fix, using same ComfyUI for SwarmUI was breaking preview feature and not anymore
Updated Windows Launchers
Both Windows launchers now:
Compose the standalone node folder and YAML config if they are missing.
Load the isolated nodes with `--extra-model-paths-config
Preserve the existing Sage Attention, Triton, auto-launch, and VRAM configuration options.
Updated Cloud Guides
The RunPod, SimplePod, and Massed Compute guides now include:
Updated June 2026 tutorial links.
A fast Hugging Face upload/download tutorial.
Current GPU, storage, template, and NVIDIA driver guidance.
Bundle 100 as the recommended broad custom-node bundle.
The automatic FoleyExtension behavior.
Correct startup commands for the standalone-node YAML config.
A reminder that Sage Attention can reduce quality or behave incorrectly with some models and can be removed from the launch command when necessary.
Moreover detailed benchmarks are made for all models and in some cases, newer Torch 2.13 brought singificant speed gains
e.g. Ideogram 4
Bundle 100
1 July 2026 V91 Update
3 new SwarmUI based presets added into presets folder:
Ideogram 4 Highest Quality - 260629.json
Ideogram 4 Balanced - 260629.json
Ideogram 4 Turbo - 260629.json
Ideogram 4 Turbo generates amazing images at 2048x2048 and really fast speed
To make them work directly get latest SwarmUI installer zip file and download Ideogram 4 Core bundle models
Make sure that you did run Windows_Custom_Nodes_Bundles_Installer.bat (inside ComfyUI installer zip file) and installed "1. SwarmUI ExtraNodes (SwarmComfyCommon & SwarmComfyExtra)"
For JSON prompt generation for Ideogram 4 model, we have ultimate new app : https://www.patreon.com/SECourses/posts/162527725
25 June 2026 V90 Update
ComfyUI-QuantOps repo updated with my fork to automatically support INT8 and Block-Wise quantized models - they are higher quality
Start commands in RunPod, Massed Compute and Windows_Run_GPU.bat edited and now includes --enable-triton-backend for even more performance
Windows_Install_Or_Update_ComfyUI.bat should be sufficient to upgrade
26 April 2026 V89 Update
New LoRA LTX2.3_Crisp_Enhance.safetensors (0.66 GB) added to the LTX 2.3 Core bundle in our model downloader app : https://www.patreon.com/posts/114517862
All LTX 2.3 presets (we have 7 so far) updated to use this new LoRA which improves quality significantly
25 April 2026 V88 Update
Upon a request of a Gold supporter I have made an amazing LTX 2.3 v1.1 Character replacement workflow
Example video if you also upvoted and leave a comment I appreciate
https://www.reddit.com/r/comfyui/comments/1svdkqa/new_character_replacement_comfyui_workflow_of/
Hopefully will make a quick tutorial video soon
This is literally the easiest to use and very best working one with so many custom features I coded
It took me more than 1 day + developing a custom node, all installed with our installers automatically
Use LTX 2.3 Core bundle to download all models automatically via our model downloader : https://www.patreon.com/posts/114517862
Use latest ComfyUI zip file ComfyUI_V88+
First run : Windows_Install_Or_Update_ComfyUI.bat
Then run Windows_Custom_Nodes_Bundles_Installer.bat and select bundle 100
Then use LTX2.3 IC LoRA Replace Any Character - 260425.json - located inside presets folder in latest zip file
Make sure to provide replacement image exactly same as first frame of the input video with only character being different e.g. cases below
This is super important and crucial for accuracy
First frame
Replacement
You can use any image model to generate such replacement characters or your photoshop skills
I used Nano Banana Pro but Qwen 2511, FLUX 2 or FLUX 2 Klein should also work fairly well
The preset automatically resize and crop your input image according to your input video
The preset automatically sets output resolution according to your input video + 720p or 1080p toggle that you can set
The preset automatically sets FPS - you can overwrite
The preset automatically uses original audio - you can turn on or off
You can determine how many seconds of the video to be used
For long video processing you can enable STREAMING CHUNK SECONDS
It will auto use last frame of previous chunk and continue
Only disadvantage is that you use same prompt for all chunks
The preset is fully optimized for lowest RAM and VRAM usage + maximum quality
Literally 1-click to install and use with such easiness and full automation
23 April 2026 V87 Update
The following presets updated to LTX2.3-Distilled-v1.1-FP8-Quant-Scaled.safetensors instead of v1.0 model
New preset LTX2.3 Video To Video 8 Steps - 260422.json added and it works great with LoRAs
I made this workflow and it auto recognize your input video dimensions, FPS and duration fully automatically
So easy to use
By using above LoRAs or even without LoRA, you can change existing videos
Play prompt and Init Image Creativity level (default 0.5) to get your desired results
Example simple prompt : Convert the video into a anime style
To get the latest LoRAs i recommend use latest model downloader and download LTX 2.3 core bundle : https://www.patreon.com/posts/114517862
Also use Windows_Custom_Nodes_Bundles_Installer.bat and install 100. LTX Audio to Video Bundle
This bundle now includes SwarmUI nodes as well
22 April 2026 V86 Update
After extensive testing on 4x RTX PRO 6000 GPUs having Massed Compute machine, new following presets added to the presets folder
To be able to use them, use our SwarmUI installer / model downloader and download FLUX 2 Klein Core Bundle and ERNIE Image Core Bundle
Multiple image input and editing becomes chaotic so i recommend using SwarmUI for this task and it works amazing
16 March 2026 V84 Update
LTX 2.3 Audio Lip synch preset significantly improved and now it is the very best available
Nowhere else has our new workflow
Now it is split into 2 different presets
LTX2.3 Audio Lip Synch Image To Video - Manual Frame Count - 260316.json
LTX2.3 Audio Lip Synch Image To Video - Auto Frame Count - 260316.json
If you use auto frame, it will auto set frame count according to your input audio lenght
Moreover, I have developed a new Custom Node to add no sound audio to beginning of videos so make lip synch perfect
So for installation, get the latest zip file, overwrite all files and run Windows_Install_Or_Update_ComfyUI.bat and Windows_Custom_Nodes_Bundles_Installer.bat and select bundle 100
Moreover, extra_model_paths.yaml file is updated to also list LTX2.3-Distilled-FP8-Quant-Scaled.safetensors in model loaders so update it if you are using SwarmUI model path
Also new preset LTX2.3 Extend Any Video With Audio - 260316.json added
It really does exten video really good but extended video audio requires some prompting
16 March 2026 V83 Update
LTX 2.3 presets added to presets folder
I have literally tested every preset out there for Audio Lip Snych and none of them were working properly
So I edited older LTX 2.0 preset and this is the best working preset atm
To be able to use these presets, first get latest zip file, overwrite all older files and then run Windows_Custom_Nodes_Bundles_Installer.bat and select option 100
Then run Windows_Install_Or_Update_ComfyUI.bat
To have all the models accurately downloaded, use latest model downloader we have : https://www.patreon.com/posts/114517862
Then download LTX 2.3 bundle it has all the models you need
You can watch this tutorial to learn how to use them but i plan a newer tutorial too hopefully : https://youtu.be/SkXrYezeEDc
Presets are set for 5 seconds 121 frames usually
You can increase their duration with like 241 frames = 10 seconds, 361 frames = 15 seconds and so on
LTX2.3_Enchance_Prompt_Feed_For_LLMs.txt significantly improved
Works great for Text to Video and Image to Video prompt generation
Upload it to your favorite LLM e.g. ChatGPT, provide input image or whatever you want and describe and tell LLM to generate / improve prompt
5 March 2026 V82 Update
Custom nodes installers made more robust and some bugs in this process fixed
6 February 2026 V81 Update
Upon request new custom node ComfyUI-CacheDiT added to Windows_Custom_Nodes_Bundles_Installer.bat
Linux/Cloud installers also has it they are synched
27 January 2026 V80 Update
Older custom node individual installers removed and now all merged into Windows_Custom_Nodes_Bundles_Installer.bat
When you run it, you will see individual nodes and now we will have bundles
Install 1. SwarmUI ExtraNodes to run SwarmUI presets
Install bundle 100 to run LTX 2 Audio To Video preset - does lip snych and more
Both RunPod and Massed Compute installer also has same node and bundle install feature, just enter numbers and hit yes to install while installing
New following presets added into Presets folder
To use the presets, use Model downloader and download necessary bundles : https://www.patreon.com/posts/114517862
LTX2_Enchance_Prompt_Feed_For_LLMs.txt added into zip file which you can upload to https://aistudio.google.com/prompts/new_chat and write your prompt and that is it. You can also upload image + this file and prompt to enhance your prompt for free
LTX 2 series - requires LTX 2 Videos Core Bundle
LTX2 Audio Lip Synch Image To Video - 261201.json
LTX2 Image To Video 8 Steps - 260127.json
LTX2 Text To Video 8 Steps - 260127.json
Z Image Base Model - requires Z Image Models Bundle
Z-Image Base - 260127.json
Z-Image Base With 2x Upscale - 260127.json
Currently Z Image doen't work accurately with --use-sage-attention so you can replace attention with --use-pytorch-cross-attention
Hopefully will make a new tutorial video tomorrow
extra_model_paths.yaml is updated and now better
New Windows_Run_VRAM_Optimized.bat has so many features to start ComfyUI with VRAM optimizations
Use this new Windows_Run_VRAM_Optimized.bat if you get OOM errors especially with Video models at long generations
To update, extract and overwrite older files, run Windows_Install_Or_Update_ComfyUI.bat and then run Windows_Custom_Nodes_Bundles_Installer.bat and select 1 and 100
13 January 2026 V74 Update
Initial support for NVFP4 LoRAs added
I tested regular LoRA on NVFP4 FLUX 2 and FLUX 1
FLUX 1 worked perfect but FLUX 2 failed
Z Image Turbo also working but I saw some quality drop
The LoRA is regular BF16 or FP8 LoRA not NVFP4
I have updated installer repo links to new ComfyUI github link
RunPod template link updated and now we fully support SimplePod which is much faster and cheaper than RunPod
SIMPLEPOD CHEAPER AND FASTER THAN RUNPOD
Now we fully support SimplePod as well please use this link to register : https://simplepod.ai/ref?user=secourses
SimplePod is faster and cheaper than RunPod and works exactly same
E.g. RTX 5090 on RunPod is 0.89 USD per hour, on SimplePod it is 0.45$ per hour,
RTX PRO 6000 on RunPod is 1.84 USD per hour and on SimplePod it is 0.79 USD per hour
Please use this template on SimplePod : https://dash.simplepod.ai/account/explore/100/ref-secourses/
For permanent storage, generate it from Storage tab with any name and size you want and when selecting template with above link, click Edit and Use, select Persistence Volume and change mount point to /workspace
Up-to-date SimplePod tutorial starting from 21:51 : https://youtu.be/yOj9PYq3XYM?si=Z86wZZLBeYzWo1Qo&t=1311
As usual follow Massed_Compute_Instructions_READ.txt and RunPod_SimplePod_Instructions_READ.txt to install and use and watch the tutorials
10 January 2026 V73 Update
New tutorial : https://youtu.be/yOj9PYq3XYM
SimplePod register : https://simplepod.ai/ref?user=secourses
We’re releasing ComfyUI installers for both Windows and Linux, built around UV for dependency installation — delivering up to 100× faster installs compared to pip.
With V73, we have moved to CUDA 13 and Torch 2.9.1
You don't need CUDA 13 installed at your system atm, you only need updated NVIDIA driver, Python 3.10 or 3.11 or 3.12 or 3.13 and Git
This brings NVFP4 and NVFP8 support
NVFP is at least 2x faster on RTX 5000 series GPUs, NVFP4 size is very small and quality almost as BF16
SwarmUI installer post (model downloader) now has NVFP4 models such as FLUX 2, FLUX 1, Z Image Turbo : https://www.patreon.com/posts/114517862
To install with V73, extract new zip file content and overwrite all previous files, then delete your venv folder inside ComfyUI and run installer
I recommend seperate install and test first
Hopefully I am producing a new tutorial for NVFP4, quality comparison and LTX 2 presets coming soon
Windows
Supports Python 3.10, 3.11, 3.12, and 3.13
Now we support a new cloud platform called as SimplePod
Works exactly same as RunPod but much faster Disk and Network speeds and much lower prices for amazing RTX 5090 and RTX PRO 6000 GPUs
Fully supports permanent network storage system but like 10x faster than RunPod
Register : https://simplepod.ai/ref?user=secourses
Our pod : https://dash.simplepod.ai/account/explore/100/ref-secourses/
Follow RunPod_SimplePod_Instructions_READ.txt to use there
Linux (including RunPod)
Supports Python 3.10
Automatic setup including Python 3.10 venv creation
Fully automated environment install
🚀 No System CUDA Needed
You do not need to install CUDA system-wide.
Only Python + Git are required (Windows), and Linux setup is fully automatic.
Make sure to have updated NVIDIA driver
🧰 Install Scripts
Windows: Windows_Install_Or_Update_ComfyUI.bat
Linux: Massed_Compute_Install.sh
Supported GPUs
RTX 1600 / 2000 / 3000 / 4000 / 5000 series
Cloud GPUs like A100, H100, B200, and more
CUDA architectures compiled for maximum coverage
7.5; 8.0; 8.6; 8.9; 9.0; 10.0; 10.3; 11.0; 12.0; 12.1 + PTX
🧩 Nodes & Add-ons (Auto + Optional)
Our installer can automatically install a large set of popular nodes, including:
Optional
ComfyUI_IPAdapter_plus
ComfyUI-ReActor
ComfyUI-Impact-Pack
SwarmUI Nodes to run SwarmUI presets
Automatic
ComfyUI-Manager
ComfyUI-QuantOps
ComfyUI-Frame-Interpolation
ComfyUI-TeaCache (latest error fixed)
comfyui_controlnet_aux
ComfyUI-GGUF
RES4LYF
🐝 SwarmUI Presets Support
We also support SwarmUI extra nodes installation, so you can run SwarmUI presets directly inside ComfyUI.
Currently available: 32 converted presets
More coming soon — including LTX 2 presets (planned)
7 January 2026 V72 Update
Fixed the broken Teacache custom node myself with forking it
3 January 2026 V71 Update
I have manually by hand converted all of our famous SwarmUI presets into fully working ComfyUI presets
This literally took hours and i tested them too
Currently we have 32 unique up-to-date presets
I plan to keep them updated as I update our SwarmUI presets from now on hopefully
Only Inpainting and Outpainting presets of SwarmUI not included since they depend on SwarmUI image editing interface
Tutorial : https://youtu.be/XWzZ2wnzNuQ
Watch above tutorial and check video chapters to learn how to do perfect Inpainting and Outpainting
Presets are located in Presets folder
Use 0-How-To-Download.txt and 0-Which_Bundles_Downloads_Which_Preset_Models.html to learn how to use them
0-Which_Bundles_Downloads_Which_Preset_Models.html is extremely useful to learn which model bundle downloads the necessary models for that particular preset
28 December 2025 V69 Update
extra_model_paths.yaml added to the zip file and it is pre-set for SwarmUI model structure
Thus by modifying base_path: F:\SwarmUI_Model_Downloader_v81\SwarmUI\ to your installation like this it will see all of your models inside your SwarmUI and you will be able to properly use them in your ComfyUI directly
Now it will auto install following nodes
ComfyUI-Manager, ComfyUI-QuantOps, ComfyUI-Frame-Interpolation, ComfyUI-TeaCache, comfyui_controlnet_aux, ComfyUI-GGUF, ComfyUI_IPAdapter_plus, RES4LYF
Extract latest zip file overwrite and run Windows_Install_Or_Update_ComfyUI.bat and Windows_Custom_Node_Install_SwarmUI_ExtraNodes.bat to install
I am working on a tutorial to show how to use SwarmUI presets in ComfyUI hopefully today
22 December 2025 V64 Update
Installers upgraded to uv pip installation
Installation on RunPod is now like 100x faster literally and many times faster on Windows and RunPod
Windows_Custom_Node_Install_SwarmUI_ExtraNodes.bat added
Automatically installs SwarmUI specific used ComfyUI nodes so that you can use SwarmUI generated ComfyUI workflows directly inside ComfyUI
RunPod and Massed Compute installers will auto install SwarmUI used extra custom nodes
30 October 2025 Update v61
Few library conflict bugs fixed
Shared requirements were not installed on Windows this issue fixed like missing diffusers library
Just overwrite files as usual and run installer
28 October 2025 Update v59
With v58 installer speed of RunPod and Massed Compute significantly improved
Now so much faster
There was a Windows installer bug fixed
Please use V59 or newer don't V58
There were also some errors on Massed Compute all fixed
Now on RunPod and Massed Compute it will ask you whether you want to install these below custom extensions or not - don't install them if you don't need them
16 October 2025 Update v57
There is a newer Sage Attention 2.2.0 and I have compiled it and updated both Windows, RunPod and Massed Compute installer
The difference is that, for example Qwen Image Edit Plus was giving black output for RTX 3090 TI with Sage Attention and not giving black output anymore
I believe this version has better support and quality so just run Windows_Install_Or_Update_ComfyUI.bat to update
Make sure to download latest zip file and overwrite older files
Here a new video that shows to to update and install properly : https://youtu.be/c3gEoAyL2IE
12 October 2025 Update v56
I have added new node RES4LYF so that now we have the some of the best Samplers and Schedulers in ComfyUI now such as beta57, bong_tangent, res_2s
Just Windows_Install_Or_Update_ComfyUI.bat and it will install these missing stuff
Fixed few library errors on RunPod and Massed Compute
RunPod and Massed Compute installers auto installs following nodes and you can remove their installation by editing Massed_Compute_Install.sh or RunPod_Install.sh before starting to install
ComfyUI-Manager - i recommend keep it
ComfyUI_IPAdapter_plus - yes you can remove if not needed
ComfyUI-ReActor - yes remove if not needed - takes time to install
ComfyUI-GGUF - i don't recommend
ComfyUI-Impact-Pack - i don't recommend
RES4LYF - i dont recommend
Removing unneeded nodes will make your install faster
On Windows we auto install only following ones others are bat files to manually install
ComfyUI-Manager, ComfyUI_IPAdapter_plus, ComfyUI-GGUF, RES4LYF
Install RES4LYF on ComfyUI will auto enable these new Samplers and Schedulers on SwarmUI as well that uses your ComfyUI backend
Just run Windows_Install_Or_Update_ComfyUI.bat and it will be auto installed
3 October 2025 Update
onnxruntime gpu errors fixed with setting onnxruntime-gpu==1.22.0
15 September 2025 Update
I have compiled Sage Attention as well on Windows
The reason is, the one I use from woct0rdho was compiled with CUDA 12.8
I compiled with CUDA 12.9 for literally every GPU out there
Just running installer will update to new Sage Attention
Tested it on RTX 3090 TI + RTX 5090 with SwarmUI + ComfyUI
Windows_Install_Or_Update_ComfyUI.bat won't ask anymore Python version if you had previously installed - only will ask first time install
13 September 2025 Update
This is a major update
The installer will now install with Torch 2.8 and CUDA 12.9, so we moved to very latest libraries
Moreover, I have compiled all the libraries that we need so it will be ultra fast to install both on Windows and RunPod (Linux) and Massed Compute (Linux)
I have compiled xFormers, Flash Attention, Sage Attention, InsightFace, Segment Anything for Windows and Linux with all GPUs support : TORCH_CUDA_ARCH_LIST=6.1;7.5;8.0;8.6;8.9;9.0;10.0;12.0
Thus this way, we literally support all GPUs starting from RTX 1000 series to RTX 5000 series including cloud GPUs like A100, H100, B200, etc
For Windows, all above libraries are compiled for Python 3.10, 3.11, 3.12 and 3.13
For Linux, compiled only for Python 3.10
I am using pre-compiled Triton (for Windows) from woct0rdho and DeepSpeed (for Windows) from official PyPi
This is a major update, I recommend fresh install but you can delete your venv and run installer again, this should work too
Moreover, kijai_Multi_Talk workflows were broken
The custom node needed for kijai_Multi_Talk was installing this node : audio-separation-nodes-comfyui
This node audio-separation-nodes-comfyui was installing a specific library and it was totally crashing the ComfyUI, it was never starting just crashing
This is why ComfyUI custom nodes can break your entire ComfyUI and make sure to test on fresh installation
I have updated Install_Multi_Talk.bat and RunPod_Massed_Compute_Instructions.txt and fixed the issue now it should work and it uses latest libraries
28 August 2025 Update v50
Kijai broken the Wan GGUF model loader
Here issue thread i made : https://github.com/kijai/ComfyUI-WanVideoWrapper/issues/1128
Therefore I have updated the kijai_Multi_Talk installer files and all of the workflows
New installer file will use 1 week older version of ComfyUI-WanVideoWrapper and all workflows updated to not merge LoRA (LoRAs not merge-able with GGUF)
21 August 2025 Update v49
New Bat file Windows_Install_Update_SwarmUI_Nodes.bat implemented
This bat file will install SwarmUI extra nodes so that you can run all SwarmUI generated ComfyUI workflows inside ComfyUI as long as you have the same model files and select them
17 August 2025 Update v48
Installers will now auto install 20250131 version of FFMPEG to RunPod and Massed Compute automatically
This is still the version that I use everywhere and working great
Moreover on Massed Compute, it will auto install RIFE and Teacache to existing SwarmUI installation
For Runpod and Massed Compute, SwarmUI auto installer will install RIFE and Teacache now. Since we already have SwarmUI on Massed Compute we don't need to install SwarmUI just update it and use manually installed ComfyUI as in tutorial
For windows please follow requirements tutorial you have to manually install
13 August 2025 Update v46
Installers for Windows, RunPod and Massed Compute and Windows update bat files are updated and made more robust
ComfyUI manager requirements were not getting installed properly and this issue fixed
I recommend to run ComfyUI update bat file
7 August 2025 Update v45
Update files made more robust to update in all cases for Windows, RunPod and Massed Compute
For RunPod and Massed Compute run installer command again and for Windows run Windows_Update_ComfyUI.bat
Now it will auto update GGUF and ComfyUI Manager as well
These 2 were not getting auto updated for some people
16 July 2025 Update v42
All MultiTalk presets updated to use new Wan 2.1 14B LightX2V CFG Step Distill LoRA V2 (T2V + I2V) (Rank 64) (0.69 GB)
SwarmUI downloader is also updated to download this file it really improves quality https://www.patreon.com/posts/114517862
15 July 2025 Update
Few workflows were using an older safetensors model file fp8, fixed with fp32
Lower VRAM configs added to MultiTalk - key is lower frame count at each chunk
12 July 2025 Update
MultiTalk workflows completelky updated and seperated into 4 folders
1: Super_Loyal_Official - which was used in main tutorial
2: Loyal_Medium_Animated
3: Less_Loyal_More_Animated
4: Lesser_Loyal_Super_Animated
Use SwarmUI model downloader bundle download to download new FusionX GGUF Q8 model which is used in new workflows
10 July 2025 Update
MultiTalk 1-click installer added to the WorkFlows
Full step by step tutorial published for this : https://youtu.be/8cMIwS9qo4M
Make sure to use SwarmUI model downloader V50, give your ComfyUI models folder and download MultiTalk bundle
How to install on RunPod and Massed Compute shown in this tutorial
Massed Compute part shows how to install and also give to SwarmUI as a backend as well so you can utilize pre-installed SwarmUI on Massed Compute with latest version of ComfyUI
Now RunPod and Massed Compute instructions will start by default with Sage Attention
15 June 2025 Update
ComfyUI requirements were missing diffusers ridiculously and i have updated our installer for Windows, RunPod and Massed Compute
Working amazing on all 3 platforms with Torch 2.7, CUDA 12.8, DeepSpeed, Accelerate, Triton, Flash Attention, Sage Attention, xFormers
Linux users can use Massed Compute or RunPod installer
5 May 2025 Update
5 May 2025 Tutorial published a must watch one : https://youtu.be/fTzlQ0tjxj0
3 May 2025 Update
Automatic download of models removed use SwarmUI Unified Downloader, you can set download path as ComfyUI folder : https://www.patreon.com/posts/118442039
27 April 2025 Update
Now the unified installer will ask you whether install with Python 3.10, 3.11 or 3.12 - you can chose - make sure that you have followed requirements tutorial and installed that Python version
All installers updated to official Torch 2.7 and CUDA 12.8
Supporting all GPUs including RTX 5000 series
Now you can install Flash Attention, or Sage attention or both - it will ask you as second step
I have compiled Flash Attention + xFormers + insightface for Windows and Linux users
Now every installer will install Torch 2.7 CUDA 12 supporting xFormers + Triton
It will support older GPUs as well
Now Massed Compute and RunPod supports RTX 5000 series + installs pre-compiled Flash Attention + Sage Attention + xFormers
Use our Massed Compute Image and RunPod Pytorch 2.2.0 (runpod/pytorch:2.2.0-py3.10-cuda12.1.1-devel-ubuntu22.04)
10 March 2025 Update
All 3 Windows installers updated
Currently installers automatically installs latest Triton (3.2+) and DeepSpeed (0.16.4), insightface (0.7.3), onnxruntime-gpu and Flash Attention (2.7.4 - I compiled for all 3 Python versions) into our VENV Windows
Supports RTX 5000 series - currently xFormers not supported so SDPA used but I am trying to compile xFormers
Clear_Triton_Cache.bat file added to automatically clear triton cache and temp folder
Run this if you get errors related to Triton package
3 February 2025 Update
New Windows_Install_Reactor.bat file added to the zip file
This bat file will install followings and its requirements automatically - run it after installing rest
It will also download necessary Reactor and all other models automatically into the models folder
Automatic installation of above added to the Massed Compute installer and RunPod installer
Also an example ReActor-Build-Blended-Face-Model workflow added to the Example_Reactor folder but i couldn't find better one - however I tested and it works really fast
3 October 2024 Update
New downloader (ultra fast) file added Windows_Download_All_IP_Adapter_FaceID_Models.bat
This file will download all of the models shared in official page automatically
Total file size is 20 GB
Official page : https://github.com/cubiq/ComfyUI_IPAdapter_plus
The commands to download all modes added to the Massed_Compute_Instructions_READ.txt and Runpod_Instructions_READ.txt files
However, still many of the presets they shared on https://github.com/cubiq/ComfyUI_IPAdapter_plus broken
I have opened an issue thread : https://github.com/cubiq/ComfyUI_IPAdapter_plus/issues/721
What I could make work are
FaceID Portrait Style Transfer
FaceID Portrait UniForm SDXL Only
Download_Models.py updated and now downloads the following models automatically (delete it before install if you don't want to download models)
sd_xl_base_1.0 (with fixed VAE)
Juggernaut-XI-v11.safetensors
RealVisXL_V5.safetensors
Hyper_Realism_V3 (SD 1.5)
Massed Compute (Recommend Cloud) :
Please register via this link : https://vm.massedcompute.com/signup?linkId=lp_034338&sourceId=secourses&tenantId=massed-compute
We have a special coupon for all GPUs : SECourses
If you want to learn more about GPUs and prices read this link : https://www.patreon.com/posts/126671823
Select RTX A6000 or Better GPU - like L40S or A6000 ADA or A100 or H100 or best is RTX 6000 PRO
Then select our image SECourses from Creator dropdown
Then follow Massed_Compute_Instructions_READ.txt
Same as my any other Massed Compute installer script
Example tutorial for learn how to install and use Massed Compute
(Starts at 12:58) : https://youtu.be/KW-MHmoNcqo?si=G1WbG-Qw4ujWvOtG&t=778
RunPod (Cloud):
Please register via this link : https://get.runpod.io/955rkuppqv4h
Then follow Runpod_Instructions_READ.txt
Same as my any other RunPod installer script
Use the template written in Runpod_Instructions_READ.txt file
Example tutorial for learn how to install and use RunPod
(starts at 22:03) : https://youtu.be/KW-MHmoNcqo?si=QN8X8Sjn13ZYu-EU&t=1323
Screenshots of Attachment Zip File
Related posts
LTX 2 Ultimate Tutorial: ComfyUI & SwarmUI on Windows and Cloud
MiniMax H3 Infinite Video: Make 2-Hour Movies in ComfyUI (0-Shot)
ComfyUI & SwarmUI on Massed Compute and RunPod: Cloud GPU Tutorial
All posts in this topic: ComfyUI: 1-Click Installer, Presets and Tutorials · New to SECourses? Start Here: Best and Most Popular SECourses Posts