This is a simple LoRA that allows you to utilize Canny and Scribble inputs as control for Anima preview 3. It is a prototype, so expect instabilities. It cna be further improved with more Control types and better tuned settings for smoother strength scaling, but all that requires some research.

It also supports Depth, but it is unstable, and suggested only as secondary input.

Special approach was taken, that allows thi model to utilize KV drop and Temporal Masking to control strength, which you can find here - https://github.com/Anzhc/Anzhc-ComfyUI-Cosmos-Reference

P.S. Don't use those approaches with other Control loras, if you are not sure if they had such training, you will encounter artifacting in case of KV, and no effect in case of Temporal Masking.

Downloads last month: 16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for CabalResearch/Anima-Canny-Scribble-Adjustable-Control-LoRA

Base model

circlestone-labs/Anima

Finetuned

(18)

this model