CatVTON In ComfyUI - Stable Diffusion Lightweight Virtual Try On Framework

[@ReedLoo-s6w]

the limitation with this is truly the resolution, after a couple tries it DOES get the garments right but the img will be low res regardless and upscaling is not doable without losing the garment details

[@TeemuKetola]

Does cat-vton have a license for commercial use and how much video ram do you need to run it?

[@bilalzidaniy5981]

i have errors such as : /CatVTON/mix-48k-1024/attention is not a folder containing a `.index.json` file or a pytorch_model.bin or a model.safetensors file where to get the from ? the repo you mentioned don't have it

[@kalakala4803]

Thanks for tutorial, looks like this one run faster

[@TheFutureThinker]

It is

[@cyberspider78910]

Is it better than OOTdiffusion ?

[@alphaaiagent]

1.20 why when I install there are these words "This action is not allowed with this security level configuration".how can i solve it

[@sirjcbcbsh]

me2 have you managed to solve it?

[@Nebiroth]

@@sirjcbcbsh Use -1 for Seed

[@GS-ny3xv]

Can I use this CatVTON in a Next.js project? Also, how to view the ComfyUI Manager?

[@fabrizioderra1233]

yes bro, just insert it in the css and ready to go

[@zoyoo6355]

Gradio link (hugging face) not working bro....

[@usserFunnySpecific]

is it possible to change the clothes in the video in the same way?

[@TheFutureThinker]

the easiest way, use DiT AI Video model to create a video using the generated image.

[@usserFunnySpecific]

@@TheFutureThinker Thanks for your response

[@ismgroov4094]

Which one is best IDM-vton or catVTon, sir❤😊

[@TheFutureThinker]

Depends

[@mauvaissigne]

Why do you work with comfyUI instead of automatic1111. I have only tried automatic1111. I always thought it was better. My goal is to make an instagram model, as realistic as possible so if you have any other tips on that as well that would be great. I really learn a lot from your vids. Thanks.

[@joselanz1699]

ComfyUI is more lightweight and is extremely customizable, giving you the opportunity to build workflows the way you want.

[@aivideos322]

Comfy is about 40% faster IMO on all renders with less vram needed. I have not touched auto1111 since i installed comfy.

[@mauvaissigne]

@@aivideos322 are the results the same or does automatic1111 give you more… accuracy?

[@aivideos322]

@@mauvaissigne Comfy is the same models. There is no accuracy, its all down to the tools that you have available. Comfy is harder to use, but more powerful and faster

[@Ton_DayTrader]

i have Error occurred when executing CatVTONWrapper: 'VaeImageProcessor' object has no attribute 'blur'

[@박은우-d2z]

me too bro,, Did you solve that problem?

[@Ton_DayTrader]

@@박은우-d2z i try new install comfyui. It work

[@krisolak2676]

Im getting some crappy results with leggings :)

[@aivideos322]

I see you are back to sam 1, :( i too have not been able to use sam 2 properly

[@TheFutureThinker]

The SAM1 custom node (not KJ edition) It have a update SAM2 loader also. But i like this node, its more accuracy for object masking. Florence 2 + Sam2 (KJ edition) is good for data analysis usage.

[@skyblaze6687]

catvton occupy almost 40-50gb vram

[@TheFutureThinker]

Woah, then you need a H100 to run it.

[@skyblaze6687]

@@TheFutureThinker dont know why the even mention 8gb isnt it look scam in github for me u can make it lighter then using just inpainting xl any other model