What is Inpainting?
Inpainting is the AI image-editing operation that regenerates a masked region of an existing image based on a text prompt, while leaving the unmasked region pixel-identical. The underlying mechanism is diffusion conditioned on both the surrounding pixels and the prompt, so the regenerated region blends seamlessly into the original. Inpainting is the workhorse operation in modern AI photography pipelines: removing distracting background objects, swapping a product's color or label, replacing a face, fixing a generated hand, or inserting a hero product into a lifestyle scene that was generated without it. Industry tools that ship strong inpainting include Adobe Photoshop Generative Fill, Krea, ClipDrop Cleanup, Photoroom Magic Edit, and underlying open models like Flux Fill, SDXL Inpainting, and Stable Diffusion 3 Inpaint. Combined with masking via segmentation models (SAM, RMBG), inpainting becomes a one-click 'fix this' tool that has largely replaced the manual Photoshop workflow in performance-creative ops.
Key statistics
- Adobe Generative Fill drove 4B+ inpainting edits in its first 12 months (Adobe Max 2024 keynote, internal usage).
- Inpainting is the most-used operation in commercial AI image editors — accounting for 40–60% of all generation calls in Photoroom, Krea, and ClipDrop analytics disclosures.
- Open-source inpainting models (Flux Fill, SDXL Inpaint) achieve 95%+ seam-quality match vs proprietary tools per FID benchmarks (HuggingFace eval leaderboards, 2025).