I wanted a colorizer that worked easily on an M1 Mac (as well as other systems). I remembered how easy it is to setup a1111, and I did not find any colorizers on the extensions list, so I implemented this one!
There will definitely be bugs so please either make an issue or hit me up on twitter.
The original repo supports this, but since Automatic1111 uses Gradio version 3.x. the only way to have a canvas would be to build a custom component that would be out of date as soon as they update to a newer version of Gradio. In Gradio 4.x+ there is a the ImageEditor component which makes things 10x easier. Will be waiting for that update.
Upload an image close in style to what you want it to look like and the model will create strokes for you resembling it.
Type descriptions and it will work
Follow the guide on how to install extensions here.
https://huggingface.co/solojungle/sd-web-ui-unicolor-pretrained-models/tree/main
Download the pretrained models (including both Chroma-VQGAN and Hybrid-Transformer) from Hugging Face:
Trained model with ImageNet - put the file imagenet_step142124.ckpt under folder framework/checkpoints/unicolor_imagenet.
Trained model with MSCOCO - put the file mscoco_step259999.ckpt under folder framework/checkpoints/unicolor_mscoco.
To use exemplar-based colorization, download the pretrained models from Deep-Exemplar-based-Video-Colorization, unzip the file and place the files into the corresponding folders:
video_moredata_l1 under the sample/ImageMatch/checkpoints folder
vgg19_conv.pth and vgg19_gray.pth under the sample/ImageMatch/data folder