Based on the sampling principles from "Differential Diffusion Animation Raw & Background Changer" workflow, this workflow differs in:

1. Reduced lots of nodes and refiner, removed all Set/Get nodes.

2. Uses AnimateDiffLightning to reduce 15 steps to 8 steps.

AnimateDiffLightning:

3. Maintains stable background and subject sampling from original workflow.

4. Uses RMBG 2.0 for subject/background separation, all nodes updated to new versions.

Operational nodes:

1. Res: Controls output resolution

2. Value1 and Value2: Control denoising levels for subject (Value1) and background (Value2). Without Controlnet control for background, higher background values cause more flickering. Subject values at 1.0 maximize prompt adherence with greater character changes.

3. Extraprompt: Defaults to prompting from original video frames; when enabled, allows custom prompting.

==

For custom prompts, enable ExtraPrompt and input through Prompt.

Choose any preferred SD1.5 model as Checkpoint.

For V2V effects demonstration, refer to the video generated using this workflow by setting subject denoise to 1.0 (same source video as Differential Diffusion Animation Raw & Background Changer).