Stable diffusion thread

[doogyhatts]

Tiktok published a new technique on depth estimation that is far more accurate than MiDaS.
https://github.com/LiheYoung/Depth-Anything

 

[billthegeek]

Still waiting for AI to generate prawn beedio.

 

[doogyhatts]

A1111 controlnet now has faceID.
https://github.com/Mikubill/sd-webui-controlnet/discussions/2530

 

[doogyhatts]

Google Lumiere is better than SVD and Pika.

 

[doogyhatts]

FreeNoise is now available on ComfyUI.
Better than FreeInit?
https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved/issues/118

 

[doogyhatts]

Generated from the DreaMoving HF site. Looks really good!
But after you split the video into frames, you can find plenty of errors.

 

[doogyhatts]

Fantastic tutorial on using the DPM++ 2S a Karras sampler.
And with an awesome thumbnail design too!

 

[doogyhatts]

FreeNoise is now available on ComfyUI.
Better than FreeInit?
https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved/issues/118

Tested FreeNoise. Faster and better than FreeInit.

AnimateAnyone is now on ComfyUI.
However, I am unsure if the public pre-trained model has sufficient data.

 

[doogyhatts]

So all SD1.5 models are cursed.

 

[doogyhatts]

New SVD model 1.1 has very windy hair.
Currently, trying out motion bucket 1.

 

[doogyhatts]

AnimateLCM is now available on ComfyUI!

 

[doogyhatts]

Tutorial on creating complete TCG cards in ComfyUI!

 

[doogyhatts]

PixAI.art now has a much better animation solution for its anime models!
Its time to unload those accumulated credits.

Anything v5 model.

 

[doogyhatts]

RAVE is for video-to-video inference. Need 12gb vram!
https://rave-video.github.io/

 

[zheng]

 

[doogyhatts]

Super Resolution upscaler for ComfyUI.
https://github.com/kijai/ComfyUI-CCSR

 

[doogyhatts]

Finally got the unofficial version of AnimateAnyone installed correctly.


Results are not fantastic at the moment. So stick to official DreaMoving HF.

 

[doogyhatts]

Stable Cascade is the next text-to-image model from Stability.AI.
https://huggingface.co/stabilityai/stable-cascade

 

[doogyhatts]

No longer needs hires fix and upscaling below 1536.
Need 16gb vram!

 

[zheng]

No longer needs hires fix and upscaling below 1536.
Need 16gb vram!

vram requirements keep going up