Does it work in ComfyUI?

#1
by juang3d - opened

I tested with apply controlnet and with diffsynth controlnet, but it does not seem to work, is there a way to make it work in comfy?

same problem

Alibaba-PAI org

We might need ComfyUI to add support for this. I'll ask about it tomorrow when I have time.

Thanks, let's hope they support it soon, I'm willing to do some testing of this, it looks great.

We might need ComfyUI to add support for this. I'll ask about it tomorrow when I have time.

thank you!

yeah~~hope this can support it

Alibaba-PAI org

If you have time, you could test this out - I just uploaded it.
https://github.com/aigc-apps/VideoX-Fun/tree/main/comfyui/qwenimage

If you have time, you could test this out - I just uploaded it.
https://github.com/aigc-apps/VideoX-Fun/tree/main/comfyui/qwenimage

Just tried, stuck at first step. Text encoder gets loaded to cpu causing Comfy to be killed (I only have 48GB RAM) so no way to see if the rest of the workflow works. Correct me if I'm wrong but comfyui's native memory logic uses gpu to load text encoder and then offloads to cpu to give space for diffusion model, which I think you should adapt to because most of us rely on maximum usage of both VRAM and RAM to run them big models today. Nonetheless thank you for following up this thread.

Sign up or log in to comment