AI Image Edit
Refine an existing frame without losing the subject. Adjust style, lighting, and details, then export or animate.
Image edit in MiniMaxVion is built for iteration. Take a generated still or upload, describe the change you want, and keep identity and composition stable while you polish the shot for shipping.
- Subject-consistent edits for campaign stills
- Fast iteration on lighting, style, and cleanup
- Hand off to image to video when you need motion
Upload a frame
Start from a generated still or an image you already have.
Describe the edit
Say what should change — lighting, background, style, or cleanup.
Generate the revision
Review the result, iterate again, or send it to image to video.
Name the change and protect the rest
A good edit prompt has two halves: the one thing that should change, and the details that must stay exactly as they are. Without the second half, the model may redraw the whole frame.
Example: "Keep the same man, the bright yellow hoodie and matching sweatpants, the white sneakers, and the pose with both hands in the pockets. Change only the setting to an outdoor seaside basketball court with hard midday sun and a sharp shadow on the concrete. Photorealistic, no text."
Common edits
- Background swap — move a person or product to a new location with matching light
- Relighting — change the time of day, light direction, or color temperature
- Style change — turn a photo into illustration, film stock, or a brand look
- Wardrobe and props — change clothing color, add an accessory, or remove clutter
- Combining references — put a product from one photo into a scene from another
Edit in small steps
When a result is close, change one category at a time: background, then light, then details. Each step stays on your board, so you can go back to any version. If the composition itself is wrong, it is usually faster to regenerate with Text to Image than to edit.
Check before you publish
Look at hands, text, logos, product edges, and reflections at full size. For product images, compare the edited result with the real item so the listing does not show features it does not have.
How is image edit different from text to image?
Text to image creates a new still from a prompt. Image edit starts from an existing frame and changes it.
Can I remove watermarks?
MiniMaxVion also offers a dedicated remove-watermark workflow in the image studio for cleanup jobs.
Can I animate an edited frame?
Yes. After editing, open Image to Video to add camera and motion to the polished still.
Can I use more than one reference image?
Yes. Most image models accept up to four reference images, for example a person, an outfit, and a location to combine in one result.
Why did the edit change parts I wanted to keep?
The prompt may not say what must stay the same. Name the protected details — face, product, label, pose — and change one thing at a time.
Pricing · Text to Image · Image to Video · 中文