Stable Diffusion with ControlNet
This model is very large, please specify its name at startup so that the model can be fully downloaded on the first launch. This way, there will be no issues when switching on the website in the future.
You can enable ControlNet to get better inpainting result when using Stable Diffusion model.
--model sd1.5 --sd-controlnet --sd-controlnet-method control_v11p_sd15_inpaint--model supported arguments:
- sd1.5
- anything4
- realisticVision1.4
--sd-controlnet-method supported arguments:
- control_v11p_sd15_canny
- control_v11p_sd15_openpose
- control_v11p_sd15_inpaint
- control_v11f1p_sd15_depth
Using local ckpt/safetensors model is also supported:
--model sd1.5 --sd-local-model-path /path/to/model-inpaint.ckpt --sd-controlnet --sd-controlnet-method control_v11p_sd15_cannyWhen using the control_v11p_sd15_inpaint method, it is necessary to use a
regular SD model instead of an inpaint model.
It should be noted that the most suitable ControlNet weight varies for different methods and needs to be adjusted according to the effect. It is recommended to start with 0.4 for canny and openpose, and 1.0 for inpaint and depth.
