Home Creators Posts Import Register Favorites Logout

Downloads

Content

Forge NEO is the up-to-date, maintained Forge Web UI; the installer sets it up with Torch 2.13.0, CUDA 13 and precompiled xFormers and Flash Attention.

1-click installers for Windows, RunPod, Massed Compute and SimplePod, with Sage Attention, InsightFace and TorchAO also precompiled. The Smart Launcher supports all GPUs from RTX 1000 to 5000 series and cloud GPUs such as H100 and RTX PRO 6000 and adds the right optimization flags.

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 installer zip file : SD_WebUI_Forge_NEO_v9.zip

  • [click here to choose a membership and Join to download zip files]

  • This installer will install latest version of Stable Diffusion Forge Web UI NEO with most performant and accurate venv setup and smart launch features

  • It is famous beloved Stable Diffusion Forge Web UI (made by lllyasviel) but fully updated to work with latest libraries

    • So this is literally Stable Diffusion Forge Web UI made by lllyasviel and it is still getting updated and maintained

  • I have prepared literally 1-click installers and GPU specific automatic optimizer ran files

  • My installer will install with Torch 2.13 and CUDA 13 + pre-compiled xFormers, Flash Attention, Sage Attention and InsightFace and TorchAO

  • My precompiled libraries are compiled with support of all consumer GPUs + cloud GPUs for Python 3.10, 3.11, 3.12, 3.13

  • Currently installer installs with Python 3.13 since it is recommended so have Python 3.13 as well in your system

  • I have developed smart_launch.py and it will be used to launch the app on Windows, RunPod and Massed Compute and it will automatically add optimization parameters according to your GPU

    • The Windows bat files and RunPod and Massed Compute scripts will use it by default so don't worry just use installer files and launch files,

    • Smart launcher supports all GPUs like RTX 1000, 2000, 3000, 4000 and 5000 series or cloud GPUs like H100, RTX PRO 6000, etc.

  • Currently scripts are set to start with disabled Sage Attention since it produces lower quality outputs

    • If you enable, enable like this : --sage2-function fp16_triton

  • To download models quickly you can use our Unified Model Downloader (updated to support this app) : https://www.patreon.com/posts/114517862

15 July 2026 Update V9

  • You need to delete your venv folder and run this update - existing installation may remain

  • I have started upgrading all of our apps into latest Torch 2.13 and CUDA 13

  • For this, I have pre-compiled the following wheels with all CUDA 13 features and with all CUDA archs : mslk, xformers, flash_attn, sageattention, torchao

  • All these libraries are properly compiled with abi3 thus works on Python 3.10, 3.11, 3.12 and 3.13

  • So I have updated the Stable Diffusion WebUI Forge - NEO installers with newest libraries

  • It will automatically generate Python 3.13 venv and install everything there, if you don't have 3.13, it will use your default python

    • Make sure to have 3.13.14 it is best

    • Forge Web UI was automatically overwriting our pre-compiled wheels with Torch 2.12 and some other inaccuratre wheels, we fixed this issue and it uses our perfect and fastest pre-compiled wheels now

  • All installers updated this way, Windows, Massed Compute (local Linux users), RunPod and SimplePod

  • Windows_Smart_Fast_Start_Recommended.bat file significantly improved and now fully detects RTX 1000 to 5000 series or cloud GPUs and start the app that way

    • On cloud GPUs (RunPod, Massed Compute, SimplePod), smart_launch.py auto used

      • Bat file is actually using that file to start

  • New supportive files added fully up to date

    • Supported_Model_List.html

    • Model_Folder_Paths.html

    • Supported_File_Formats.html

Windows Requirements

  • Python 3.13.14, FFmpeg, CUDA 13 cuDNN 9.17, Git, Visual Studio Community Edition with Desktop Development with C++ (all checkboxes checked)

    • Make sure that your NVIDIA driver is updated (min 590+)

  • If you get any errors follow below video and its source link

  • Just use Windows_Install_or_Update_Forge_Neo.bat to install, after auto install it will auto start the App

  • Later just use Windows_Smart_Fast_Start_Recommended.bat to start fast

  • To install extensions use Windows_Slow_Start_Use_This_For_Extension_Install.bat

Massed Compute or Local Linux (Recommend Cloud) :

  • Follow : Massed_Compute_Instructions.txt

RunPod & SimplePod (Cloud):

  • Follow : Runpod_SimplePod_Forge_NEO_Instructions_READ.txt

Screenshots

  • App main interface

  • Auto installed working ADetailer (automatically detect faces or persons or hands and inpaint)

  • Auto installed working ControlNet

  • Never OOM feature and Image Stitch

  • Spectrum Integrated and Torch Compile

  • Auto Installers Files

  • Smart Launcher - auto sets performance arguments according to your GPU model

15 July 2026 Supported Model Families

15 July 2026 Supported File Formats

Model Folder Paths

Related posts

All posts in this topic: AI Image Generation and Editing Apps · New to SECourses? Start Here: Best and Most Popular SECourses Posts

Files

Comments

DanO..

Fantastic News! I know ComfyUI is great for coders and developers but it SUCKS to use! A fucking mess. Thanks for this. It's been what I've been waiting/hoping for.

Nicolas Meunier

Fantastic, but to prevent having all models in several version, how to change the model folder with your smart_launch.py? the traditional way by adding arguments in web-user.bat is not working here... or I am doing something wrong

SECourses

it is literally same just add args to bat file . read the post you will see it

SECourses

yes i saw ipndm missing. i dont know exactly but this is what they have

SECourses

yep this one is great and looking promising

Nicolas Meunier

if I launch manualy Webui-user.bat it works (with my arg in it) but if I use your launcher script, it does not, and I dont see, where you are calling webui-user.bat in your script

SECourses

dont manually launch. i made 2 bat files : Windows_Slow_Start_Use_This_For_Extension_Install.bat and Windows_Smart_Fast_Start_Recommended.bat

Nicolas Meunier

I know, but your 2 bat files dont use webui-user.bat, so they dont use my args.

SECourses

my bat files are passing any args you give. that is how i am also giving some args. when you open you will see

Nicolas Meunier

well I think we are 2 not really listening, but I promise I have read and open all files and now understand what you mean. In fact in your own page about the models on github, you are writing that we must change the arg with path in the bat file W"ebui-user.bat" but in fact here you are saying that we have to change that in the launcher bat file. My confusion came from that. All is set correctly now, thanks for your work.

Kevizz Men Of The World Art Adult/18+

Neo works great and fast. Thank you. I added ADetailer, ReActor and Flux Kontext Image to Image Editing to the extensions and they all work.

Frédy Nativel

hello, how to use the dark theme correctly. I add --theme dark to set COMMANDLINE_ARGS in webui-user.bat. Is the the right way? if i launch Windows_Smart_Fast_Start_Recommended.bat, that's no apply the dark theme

SECourses

i think if you set your windows theme to dark it auto does. also try this args : --theme dark

Frédy Nativel

hello, how to use wan 2.2 in Forge NEO ?, what are the clip, vae, models supported please

SECourses

i recommend to download Wan 2.2 Core 8 Steps Bundle (Total: 70.41 GB, 11 models). when you look at it you will see it has Wan 2.2 VAE (1.31 GB) and UMT5 XXL FP16 (Default for SwarmUI) (10.59 GB) . if i recall correctly wan 2.2 image to video uses wan 2.1 vae

Hoàng Giang Sơn Trương

Hi dear Furkan. Whats the difference between these WebUI Forge - Classic and this Neo version?

SECourses

WebUI Forge - Classic = automatic1111 Neo is Forge Web UI with newer features like wan 2.2 support

Roger Kaufman

In Forge Neo, the sage attention flag is --sage instead of --sage2-function fp16_triton like Forge Classic

Phillip Bopp

Either I am missing something and I can't get it to work . I have all of my models, LORAs and other items saved in a different folder. How do you we get it to work where Forge and Comfy redirect to that central folder instead of having multiple copies of some very large model files.

Tyler

I've been using the same instace of ForgeUI for... 8+ months now? Will your version here run faster than the one I'm still on? Is Flash Attention still fast, but not as fast as Sage Attention, but there is no Sage Attention for ForgeUI*?

Tyler

Please clarify, does sage attention work out of the box? What is the full command line arg to add to get it to work? I assume once this is applied, this is the fastest current possible configuration for forge gens?

SECourses

we have both. i recommend make a fresh install and compare each one. fresh install into a new folder. my installation wont affect your other one

Kay Tichelmann

suddenly not working anymore, even not with a clean install :(

Fernando Martin

Hi Dr. it seems they are very active in the project and there was a major update which your version doesnt support and stopped working and/or fails to update. if you can fix it will be great news for the ones who use Forge. As always, thanks for your work

SECourses

3 Feburary 2026 Update V2 Installers upgraded to Torch 2.9.1 and CUDA 13 I also upgraded the part we install libraries to uv thus much faster Tested on Windows, RunPod and Massed Compute with a fresh install and with FLUX Dev model and works perfect Make sure to use updated NVIDIA driver on Windows New RunPod template : https://get.runpod.io/SECourses_CU13 Now we support SimplePod too read Runpod_SimplePod_Forge_NEO_Instructions_READ.txt We always install latest version of Forge Web UI NEO

Kay Tichelmann

Hi,sorry, i was absent a while because of work...thanks alot for the new Installer, the Python 3.11 (now mandatory) did the trick :)

Josh S

I'm curious why you have it disable sage on a 5090 RTX GPU? Isnt sag considered faster and newer?

SECourses

it causes so many errors and black outputs with this app but you can certainly enable it. just edit Windows_Smart_Fast_Start_Recommended.bat and remove --disable-sage

Josh S

thanks for the response! I actually was playing with AI to help me tweak your neo install. It had me use these commands (--highvram --bf16-unet --bf16-vae --bf16-text-enc --cuda-malloc --pin-shared-memory --fast-fp16) and it is definitely faster. Honestly I have no idea how they change things other then shaving a good 20 seconds off my 4k images I'm rendering

Josh S

Also I noticed Kohya HRFix Integrated is missing from this version.

SECourses

you are welcome thanks for comment

Fossil Bluff

I've been an A1111 user forever and now moving over to this. I'm instantly stuck on trying to configure the paths for my models. I used to enter them into the webui_user.bat but now this new launcher doesn't allow me to modify the paths. Is there a new method?

SECourses

hi easy. edit the Windows_Smart_Fast_Start_Recommended.bat file. and add args and full paths . e.g. : --ckpt-dirs "F:\SwarmUI_Model_Downloader_v81\SwarmUI\Models\Stable-Diffusion" few examples : --ckpt-dirs, --lora-dirs, --vae-dirs, --text-encoder-dirs, --embeddings-dir you can also give full new homes : --forge-ref-a1111-home, --forge-ref-comfy-home, --forge-ref-comfy-yaml

SECourses

ALSO --model-ref is the main switch (set in parser and used as the models base path):

Roger Kaufman

Thanks! --model-ref seems to take care of the other directory paths. This should probably be noted in the install instructions.

Liam Gilroy

Is there a reason you aren't going to Torch 2.10 yet? I notice that the base Forge Neo installer now automatically goes for 2.10cu130, but I know some of the other tools take time to update.

SECourses

torch 2.10 is still have problems with many libraries. i am waiting it to mature. 2.9.1 working great no need to upgrade at all

sygilWLF

howdy, the new version, (v5) seems to not work with Adetailer. is there a way i can fix this? (i have 0 coding xp) if you need the error report, dm me and ill send it to you

SECourses

just fixed for you To use adetailer Step 1 : Start with Windows_Slow_Start_Use_This_For_Extension_Install.bat - delete older adetailer extension if you have before starting Step 2 : Install from URL : https://github.com/FurkanGozukara/adetailer_Forge_NEO Step 3 : Restart and start again with Windows_Slow_Start_Use_This_For_Extension_Install.bat - it will download models and will work

sygilWLF

thank you so much! i appreciate you.

Christoph Behrmann

better to have 3.13 ? for better or best results? -- This program is tested with 3.13.12 Python, but you have 3.11.9. If you encounter any error regarding unsuccessful package/library installation, please downgrade (or upgrade) to the latest version of 3.13 Python, and delete the current Python "venv" folder in WebUI's directory.

SECourses

no not better. our libraries are also compiled for 3.11 not for 3.13 atm. 3.11 working perfect i didnt see any error.