Inpainting

The classic method uses a mask: a black-and-white layer that marks the area to change. The model keeps the pixels outside the mask, adds noise to the masked area, and then denoises it from your description, taking care that the new content blends with what surrounds it in light, color and texture. A small area on a clear background usually works well. A large area, or one that crosses a face, lettering or a complex pattern, is harder, and the edge of the mask can leave a visible seam.

Typical uses are practical. A photographer removes a passer-by from a landscape, a shop owner replaces a cluttered table behind a product with a clean surface, an illustrator fixes a hand that came out with six fingers, and a designer swaps the color of a jacket in a portrait of a fictional character. The related technique called outpainting does the opposite and extends the canvas past its original borders. Both depend on the same idea, the model filling in a gap with plausible content.

Newer editing models skip the mask. You describe the change in plain words, such as make the sky overcast or remove the bicycle, and the model works out where to act. This is easier to use, at the price of less precise control over exactly which pixels change, so check the whole image afterward, not only the edited zone. Whatever the method, save the original: an edit is a new image, and the model can alter details that you did not intend to touch.

One more tip concerns the prompt. When you describe an edit, name both the change and what must stay, for example replace the red mug with a glass of water, keep the table and the light unchanged. Short, concrete instructions beat long poetic ones, and a single edit at a time is easier to judge than three at once.

Resolution matters too: very small regions leave the model little context to match, so zooming in on the area before editing often gives a more convincing blend.

NeverNo edits images by instruction rather than by painting a mask. In the Studio, editing by instruction is one of the tools the image generator page lists for paid plans, next to upscaling and background removal, and the API has a matching endpoint, POST /image/edit. You give an image and a plain description of the change, and the included editing engine is FLUX.2 klein from Black Forest Labs. A premium engine, GPT Image 2.5, can be used instead and is billed in credits.

Every image you send is checked against NeverNo's hard limits before the edit: nothing involving minors, and no images of real, identifiable people without their consent. The pictures you make or edit are not kept in a gallery on our servers: they are stored encrypted in your browser, and the API returns images for you to save yourself. There is no brush or mask tool in the Studio at this time.