What problem does it solve? Removing objects, watermarks, or blemishes from images and replacing masked regions requires precise local edits that generic image generators cannot target. This Skill routes mask-driven inpainting jobs to the right RunComfy model and falls back to description-based editing when no mask exists. ## Core Features & Use Cases - Mask-Driven Inpainting: Sends a source image plus a grayscale mask (white = inpaint, black = preserve) to Z-Image Turbo Inpainting with tunable strength and control-scale parameters. - Description-Based Fallback: Routes to Nano Banana 2 Edit, GPT Image 2 Edit, or FLUX Kontext Pro when no mask is available and the region must be described in spatial language. - LoRA Style-Locked Inpainting: Supports the Z-Image Inpainting LoRA endpoint for brand-consistent fills. - Use Case: Remove overhead cables from a street photo by supplying the image URL, a mask URL covering the cables, and a prompt like "preserve rooflines and sky gradient" at strength 0.5. ## Quick Start Ask the agent to remove a specific object or watermark from your image by providing the image URL and, if available, a mask URL, and it will run the appropriate runcomfy inpainting command.
