Inspired by https://civarchive.com/models/2443202/anima-canny-control-lora-controlnet-like

The training images were processed using PiDiNet Line preprocessor; when generating images, you can use PiDiNet Line preprocessor or another scribble preprocessor. Please try out the effects of different preprocessors for yourself.

The workflow has been embedded in the image. Here is the link to the custom node used in the workflow: https://github.com/Mirumo0u0/ComfyUI-Cosmos-Reference