Home Creators Posts Import Register Favorites Logout

Downloads

Content

You can find the most recent version here:https://www.patreon.com/posts/surprise-update-150848984

Hello my friends,

The new installer version is here! I’ve updated it to fully support LTX 2 and Flux 2 (Klein 4B + 9B).

And that’s just the appetizer.

LTX 2 Workload comes with four workflows. I made them as easy as possible.

Text → Video

Image → Video

Text + Your Audio → Video

Image + Your Audio → Video

That’s right — you can drop in your own voiceover, sound design, whatever you want It’s pretty good at lip sync too. The creative freedom is actually stupid right now.

🎥 So where’s the damn tutorial video?Already recorded. Currently in final cut. You know me, I am a perfectionist, that’s why it took a little longer this time. Youtube Videos about AI Video models usually do take quite some time and LTX 2 gave me some headaches sometimes to be honest.

⚙️ Hardware & Performance (RTX 5090 Benchmarks)These workflows are powerful, so keep an eye on your VRAM. Here is what I’m hitting on my 5090:

1920 x 1088: approx. 10 seconds of video.

1280 x 768: approx. 15 seconds of video (before hitting OOM).

!The Number of Frames need to be dividable by the FPS + 1!So if you set it to 240 Frames add +1 (10 second Video)The resolution shall be dividable by 64. So 640 x 640, 1920 x 1088,  1280 x 768

So maybe start with 640 and 5 seconds and work your way up until you are out of VRAM:

💎 Exclusive Content:For my AI Art Collectors and AI Art Architects, I have prepared a collection of best-practice examples that will drop alongside the video.

🚨 Important Troubleshooting TipIf your video gets "stuck" at the beginning of the Image Workflows:Activate the LoRA Loader and use the Static Camera Workflow at a strength of 1.0.

The video drops very soon. Get your installers ready!

When installing ComfyUI use the UV Dependency resolver it is still marked as Experimental, but it will become the default in the next major version.

Best Regards

Chris

P.S.: If you run in this error or the one below:

Decativate the preview by marking them and pressing crtl + B

This is a related to a current bug in ComfyUI.

Files

Comments

cortlandt parker

hi. this only works with comfy-ui? Can it be installed as a standalone program such as fooocus or framepack?

David Thielen

On the ComfyUI subreddit most everyone says Wan 2.2 is the best for uncensored Txt2Vid. But you clearly prefer Flux. In what ways is Flux superior to Wan? I'm not disagreeing, just trying to learn.

cortlandt parker

thank you for the response. I was not commenting on what was or is better. The question was if this can be run with out comfy ui. for example i can go to github and find fooocus and it automatically intalls without installing or requiring comfy ui or pinokio. framepack can also be installed without installing comfy ui.

AI Knowledge Central

Hi, sorry for replying so late. I don't get informed via mail when someone comments, only when i get a direct message. Currently my installer Supports the following stand allone installer: ComfyUI ForgeUI and AI-Toolkit for Lora training. You can pick a Workload for ComfyUI and it will download all the the models, ComfyUI Custom nodes and will download and install my workflows

AI Knowledge Central

Hi, To be honest it depends. Sometimes you want to generate a specific Image and then you dont have the loras etc for flux. For Images I love to use Z-Image. For different Poses and Editing my Favorite is Qwen Image Edit. For Text 2 Videos LTX 2 is great but Wan 2.2 clearly has much more Loras currently. So sometimes you have to pick the tool not just because its the best, but because the tool has what you need to achieve your desired result.

David Thielen

It’s too bad we can’t make a LoRA that works with all models.

Jgrov

I am getting this error on a fresh install using the automatic installer, did the whole thing.. "Error while deserializing header: incomplete metadata, file not fully covered"

Jgrov

Sure I will dm you..I got it to run actually..not all the files were downloaded/ but comfyui crashed during generation, just text to video..I have a4500 with 20gig vram