Simple video to painting conversion (Animatediff+LCM)

Simple ComfyUI Video Rework Workflow:

5. The workflow adds Impact facial refinement, which ensures normal facial features when generating videos where the body or facial area of the character is relatively small.

1. Use Animatediff for video reworking, where the length of context can be chosen according to the configuration.

2. Use LCM sampler with LCM dedicated LoRa to achieve better results in fewer steps. The recommended sampler CFG is 1-3, and the recommended steps are 4-8.

3. Use segment anything semantic segmentation for image cutting, for local retouching (it is recommended to deactivate this module when not in use to improve output efficiency).

4. In this workflow, Controlnet uses lineart to control the visuals. It is also recommended for everyone to use Tile+openpose to preserve the original video's movements and colors as much as possible.