Tool · Photo editing
Edit a photo by describing what you want
Removing a passer-by, changing a sky, swapping an object: classic retouching takes layers and time. Here you paint the area, describe the result, and the model does the rest — or you describe the change without painting anything at all.
Ways to edit, then compare
Erase
Brush over what is in the way — an object, a reflection, someone in the background. The engine (Bria Eraser) removes it and rebuilds the background behind.
Replace an area
Paint an area and describe what should appear there. The Inpaint tab offers nine engines, all kept because they respect the area: GPT Image 2.6 Flare (the default) and Sunburst, GPT Image 2.5 Flare and Sunburst, Ideogram 4.5, Bria Fibo 2.0, Qwen Inpaint, Z-Image Turbo and Bria GenFill. For GPT Image, the area is shown to the model, then the server pastes the original image back all around it, with a softened edge (except for an image more elongated than 3:1, or if the paste-back fails: you then get the model’s image as is). GPT Image renders up to 4K; the other engines keep the image size.
Edit the whole image with a sentence
No mask: the editing models in the catalogue — Nano Banana, GPT Image, Qwen Image, FLUX 2 Pro, Hunyuan Image — take an image and a sentence and apply the change.
Edit an image with ChatGPT, or with its model here
GPT Image is OpenAI's image model — the one most people know through ChatGPT. Here you use it without a ChatGPT subscription, in two ways. Without a mask, you give it the image and a sentence and it edits the whole image. On one area, in the Inpaint tab, GPT Image 2.6 and 2.5 see the area (circled for 2.6, highlighted for 2.5), and the server pastes the original pixels back outside it, so the background and the rest of the subject stay those of the original. And if you work in ChatGPT, the Beemm Vision MCP connector brings four edits there — upscale (×2 or ×4), background removal, filling an area with FLUX Fill, frame expansion — from the conversation; the Inpaint tab’s engines and the paste-back stay in the app.
Change the light or the camera angle
Two tabs change the shot with no prompt. Relight redoes the lighting: you place the light on a 360° sphere or with a quick position (Top, Front, Right, Left, Back, Bottom), then pick Soft or Hard, the brightness and the color (the Relight guide). Angles moves the camera around the subject: rotation from 0 to 360°, tilt from -30° to 90° and zoom (the Angles guide).
Compare before you keep it
The result is added without replacing the original. In a workflow, the Compare node puts before and after side by side, or two engines on the same edit.
How to edit a photo
With a mask or without, depending on the precision you need.
- 1 · Open the imageFrom a generation preview, a workflow gallery or a board.
- 2 · Pick the moveErase to remove, Inpaint to replace an area, Relight for the light, Angles for the viewpoint. For an overall change, hand the image to an editing model instead.
- 3 · Paint and describeThe brush marks the area; for Inpaint, a sentence says what should appear, and you pick the engine in the menu (GPT Image 2.6 Flare by default). The cost is shown before you run it.
- 4 · Check and keepCompare with the original. In Generate, the edit’s tile appears as soon as you click and survives a page refresh. On a board, the edit becomes a new version of the element; the original stays available.
Frequently asked questions
- What is the difference between erase and inpaint?
- Erase removes what you paint and rebuilds the background, with no prompt. Inpaint replaces the painted area with what you describe.
- Can I edit an image with AI without drawing a mask?
- Yes, with an editing model from the catalogue: supply the image and a sentence — "replace the background with a beach at sunset" — and the model applies the change.
- Does AI photo editing leave the rest of the photo untouched?
- Erase and Inpaint target the painted area, and the rest depends on the engine. With GPT Image 2.5 and 2.6, the server pastes the original pixels back outside the area. Ideogram 4.5 works in high precision, moving nothing outside the mask, and Bria Fibo 2.0 leaves the pixels outside the mask untouched. A prompt-only edit, with no mask, reinterprets the whole image. Either way, compare with the original before delivering.
- Can I edit an image with ChatGPT's model?
- Yes: GPT Image, OpenAI's image model that most people know through ChatGPT, is in the catalogue, with no ChatGPT subscription. It edits the whole image from a sentence, or a single area in the Inpaint tab, where versions 2.5 and 2.6 are pasted back outside the area.
- How much does an AI photo edit cost?
- Each edit uses credits depending on the engine, and the cost is shown before you run it. Details are on the Pricing page.
Edit a photo
Open the studio, open an image in the editor, and pick Erase or Inpaint.
Keep reading
- Guide: Edit a photo with AI, by painting or by describingErase an object, replace an area, expand the frame, upscale up to 16× or remove the background: AI photo editing in 4 steps, cost shown first.
- Guide: AI inpainting: change one area of an image without touching the restAI inpainting: paint an area, describe the change, and only that area changes. GPT Image 2.6, Ideogram 4.5, Fibo 2.0: which one to pick.
- Guide: AI relight: change the direction and color of the light in a photoAI relight: choose where the light comes from, its color, soft or hard, with no prompt to write. Beemm Vision's Relight tab, step by step.
- Guide: Change the camera angle of a photo with AI, from a single imageChange the camera angle of an image with AI: orbit 360° around the subject, high or low angle, zoom, with no prompt. The Angles tab, step by step.
- Nano Banana on Beemm VisionUse Google's Nano Banana 2.1, Nano Banana 2 and Pro image models on one canvas: legible in-image text, editing from up to 14 references, up to 4K.
- GPT Image on Beemm VisionOpenAI's GPT Image 2.6 (Flare, Sunburst), plus versions 2.5, 2 and 1.5, without a ChatGPT subscription: generation, editing, area edits, eight formats.
- FLUX on Beemm VisionUse FLUX 3 from Black Forest Labs: images up to 4K, FLUX.3 Klein and Klein Base, 5 to 20 second videos at 1080p with a draft mode, and FLUX.2 Pro.
- A logo in vector, not an image to redrawCreate a logo with AI and get it as a vector SVG with Recraft, or work the lettering with Ideogram, then compare the options side by side.
- Remove a background in one clickOnline, automatic background removal: the AI isolates the subject as a transparent PNG. Portrait mode for people, Matting mode for hair and fine edges.
- The AI image generator that does not lock you to one modelAn AI image generator with the whole catalogue in one picker: compare models on the same brief, edit, upscale, on a single credit balance.
- The full catalogueEvery image and video model available in the canvas.