Kakigōri Maker

impl in https://github.com/Mikubill/sd-webui-controlnet/pull/262

Looks nice. PR welcome or will implement later if no one does

Nice idea! But two questions: 1. Are original models trained by lllyasviel tough enough to support such large resolution? 2. Currently we used `F.interpolate("nearest")` to handle Hires.Fix stuffs, not a...

颜色控制可能还是训练一个能支持色彩输入的Control Net更好… 不过物体位置标识应该可以借助 Paint with words/gligen/universal-guided 等来实现 fyi: https://github.com/Mikubill/sd-paint-with-words

~~Are you using images that have a black background and white lines? Canny model requires a white background and black line image as input (currently if preprocessor is none).~~ Not...

Since model could be split into base_model(~4G) + ControlNet_model(~1.6G/700M), extension itself only reads the ControlNet part and WebUI will handle the rest. On lowvram/medvram mode WebUI will read base model...

Compatible, just match and use.

Hand Pose Estimation is disabled by default, you may need to enable it manually. https://github.com/Mikubill/sd-webui-controlnet/blob/fca33937af7bf45dac94616531643ce228c88d6b/scripts/processor.py#L69 (Maybe we should add a setting tab to configure such things)

Hmm, any screenshots?

Also close #127 #129. Thanks for your contribution!