comfyui extension more features! :smile:

tencent-ailab / IP-Adapter

The image prompt adapter is designed to enable a pretrained text-to-image diffusion model to generate images with image prompt.

Apache License 2.0

4.95k stars 319 forks source link

Open cubiq opened 9 months ago

cubiq commented 9 months ago

hey @xiaohu2015 it's been a while :smile: I thought you might be interested in the new features of the comfyui extension.

First one is integrated timestepping

timestepping

Second is compatibility with animations!

Congratulations on your amazing models, the best thing that happened to Stable Diffusion :smile:

Are you going to release any new model soon?

xiaohu2015 commented 9 months ago

@cubiq it is very cool, thanks a lot.

as for the new model, I only have some preliminary ideas. If there is any new news, I will tell you as soon as possible.

cubiq commented 9 months ago

as for the new model, I only have some preliminary ideas. If there is any new news, I will tell you as soon as possible.

have you considered LCM?

xiaohu2015 commented 9 months ago

as for the new model, I only have some preliminary ideas. If there is any new news, I will tell you as soon as possible.

have you considered LCM?

it seems current model can be compatible with lcm-lora

cubiq commented 9 months ago

yes, it works, I was wondering if it could be further optimized or if it wouldn't make any difference to train for it

xiaohu2015 commented 9 months ago

yes, it should be. But compared to LCM, I would like to train ip-adapter with sdxl turbo https://stability.ai/news/stability-ai-sdxl-turbo

cubiq commented 9 months ago

that would be amazing!

Changlin-Lee commented 9 months ago

yes, it should be. But compared to LCM, I would like to train ip-adapter with sdxl turbo https://stability.ai/news/stability-ai-sdxl-turbo

Looking forward to seeing it! Any progress on ipadapter with sdxl-turbo?