inpainting-regions

Compiles a Visual Intent Object into a ComfyUI inpaint prompt JSON using an SDXL LoadImage+VAEEncode+SetLatentNoiseMask template. Use whenever the planner emits a step with goal=inpaint, or when the u

SKILL.md

GitHub

inpainting-regions

Template+patch Skill for masked inpainting.

When this Skill applies

  • Planner step with {goal:"inpaint"}.
  • Caller has an input image and a mask PNG.

Intent overrides

VIO Node.input
prompt 6.inputs.text
negative_prompt 7.inputs.text
inputs.image 10.inputs.image
inputs.mask 11.inputs.image
params.steps/cfg/sampler_name/scheduler/seed 3.inputs.*
params.denoise 3.inputs.denoise
model.checkpoint 4.inputs.ckpt_name

Scripts

  • scripts/compile.py — VIO → inpaint prompt JSON.