AI Image Editor
Change anything in a photo by describing it: remove an object, swap a background, restyle, extend the frame. Runs on Flux Kontext and Nano Banana models.
What can you create with AI Image Editor?
AI image editor driven by a prompt
Describe the change in plain language and the model makes it. 'Remove the person on the left', 'change the jacket to olive green', 'make it look like late afternoon' — no selections, no layers, no masking. It is dramatically faster than a traditional editor for the kind of change that is easy to say and tedious to execute, and it is worse at anything requiring precision to the pixel.
Edits it handles: remove, replace, restyle, extend
Four operations cover most of what people actually want. Removing an object and having the background rebuilt behind it. Replacing something — a garment, a sky, a background — with something else. Restyling the whole image while keeping the composition. And extending the frame beyond its original borders, which is useful when a crop is too tight for the format you need.
Keeping the untouched parts untouched
The failure mode to watch for is collateral change: you ask for one thing and the model quietly re-renders a face or shifts a colour elsewhere. Naming what must stay the same helps — 'change only the background; keep the person, the lighting and the framing exactly as they are'. Flux Kontext is noticeably better than general models at honouring that kind of constraint.
Flux Kontext vs Nano Banana for edits
Flux Kontext was built for instruction-driven editing and is the default choice when the edit is surgical and the rest of the image must survive untouched. Nano Banana 2 is faster, cheaper and better at conversational iteration — you can follow up with 'a bit warmer' and it understands the reference. For a first pass, Nano Banana 2; for a careful one, Flux Kontext.
Editing photos for commercial use
Paid plans include a full commercial license covering edited images, which matters for product photography, marketing assets and client deliverables. Licensing covers our output, not your input — an edited photo of something you do not have rights to is still a photo of something you do not have rights to, and editing does not change the underlying permission you need.
What usually goes wrong, and how to avoid it
The failure that costs the most time is silent collateral change — a face subtly redrawn, a colour shifted, a logo garbled — because you do not notice until the asset is out. Name what must stay identical in every edit prompt, then compare against the original at full size before shipping. Precision work at a specific pixel boundary is still a job for a conventional editor, not for a prompt.
Checking an edit before you ship it
Flip between the original and the result at full size and look specifically at three things: faces, text, and brand marks. Those are what a viewer checks first and what the model is most likely to have quietly rewritten while doing what you actually asked for. A thirty-second comparison catches almost every collateral change, and it is far cheaper than finding it after the asset is live.
How to use AI Image Editor
- 1Sign in (free)
Create a SoonMaker account in one click with Google. You'll get free credits immediately — no card required.
- 2Upload the image you want to fix
Drop in the photo you want to run through AI Image Editor. You don't need any editing skills — there are no layers or masks to set up.
- 3Let the AI process it, then download
The AI does the work automatically. Compare it against your original and download the finished file.