Inpainting is the AI image-editing capability where you mask a specific region of an image and ask the model to regenerate just that region while keeping everything else unchanged. It is the bridge between "the model generated something close but not quite right" and "the final asset I can ship". Inpainting is built into most production image tools in 2026 — Photoshop's Generative Fill, Midjourney's Vary Region, Stable Diffusion's mask mode, FLUX inpainting variants.
How inpainting works:
- The user paints a mask over the region to change.
- The model receives the original image, the mask and a prompt describing what should appear in the masked region.
- The model regenerates only the masked pixels, conditioned on the surrounding context so the result blends seamlessly.
Common production use cases:
- Object removal — remove tourists from a hero shot, remove logos from stock photos, remove distractions from product images.
- Object replacement — swap a t-shirt colour, change a sky, replace a piece of furniture.
- Adding objects — drop a product into a lifestyle scene, add a person to an empty environment.
- Fixing flaws — repair AI-generation artefacts (extra fingers, weird hands, distorted backgrounds).
- Local style changes — apply a different texture or style to a specific region.
The 2026 tools that lead:
- Adobe Photoshop Generative Fill — the consumer-grade king; integrated into every designer's toolkit, generally licensed for commercial use.
- Midjourney v7 Vary Region — built into the Midjourney web app; quick and aesthetically strong.
- DALL-E 3 / GPT-5 image edit — conversational inpainting inside ChatGPT.
- FLUX Fill / Stable Diffusion inpainting — open-weight options with full control via ComfyUI.
- Krea, Magnific — third-party tools specialising in restoration, upscaling and detail enhancement.
The companion technique is outpainting — extending an image beyond its original boundaries. Same model, different mask: instead of masking a region inside the image, you mask the new canvas area outside it. Useful for changing aspect ratios, adding negative space for typography, or extending a scene for cinematic crops.
For a US team in 2026, inpainting is the technique that turns image generation from "roll the dice" into "production pipeline". The first generation gets you 80% of the way there; targeted inpainting cleans up the remaining 20% (problematic hands, weird typography, off-brand colours, unwanted elements). Mature creative workflows now include inpainting as a standard finishing step, often automated — a small model identifies likely artefacts and a larger model fixes them, all without human intervention until final review.
