AI Photo Editor for Edits You Can Describe

How editing a photo with a text prompt works

Edit an image by describing the change you want, or generate a new image from a prompt. Choose the available model, output size and format, then review the result before downloading. Specific instructions about the subject, lighting and areas to preserve make the intended edit clearer. AI can alter details unexpectedly; check text, faces and product features in the output.

Cost: Free options are available. Pro options or use beyond the free allowance may require credits; see the guide below.

How to edit or create an image with a prompt

Choose Edit or Create
Edit is selected by default and changes an image you supply. Create makes a new image from the prompt alone and needs no upload.
Pick the image (Edit only)
Click Upload, or open Files to reuse an earlier upload or a result from another tool. JPG, PNG, WebP, HEIC, HEIF, AVIF, BMP and TIFF files up to 200 MB are accepted.
Replace the sample prompt
The Prompt box starts with the sample text "Improve lighting, sharpen details and keep the photo natural." Overwrite it with your own instruction. In Create mode the button stays disabled until the box contains text.
Run the request
Click Generate / Edit with AI. The request waits in a queue on our servers and is then passed to OpenAI's image model. A request is stopped after 15 minutes.
Compare, download or refine
After an edit, drag the slider across the preview to compare Original with the result, then click Download. The result is also saved in Files, so you can select it and apply a further edit on top of it.

Edit, Create and the prompt

Apart from the image, the tool has two inputs: the mode and the prompt. There are no controls for size, quality or file format; the image model decides those.

Edit: Sends your selected image together with the prompt. The server instructs the model to follow your request and preserve the details you don't ask to change, so you don't need to add "keep the rest unchanged". The model still redraws the whole picture, so check faces, small text and logos in the result.
Create: Sends only the prompt, so no upload is needed. Use it for new scenes, illustrations or mock-ups.
Prompt: Plain sentences describing the change or the scene. The tool sets no length limit, but every click is a separate request, so a precise first prompt saves requests.
Output size and format: Not adjustable. The model chooses the resolution and aspect ratio, and the file is normally a PNG. For exact pixel dimensions, use Resize Image afterwards; for more pixels, run the result through AI Upscale.

Prompts that give predictable results

These are rules of thumb for instruction-based image models, not guarantees.

Name the change and where it happens

"Replace the grey sky behind the building with a clear evening sky" gives the model a target and a location. "Make it nicer" leaves both open, and the model fills the gap with its own taste.

One change per run

Five edits in one prompt make it more likely that one is skipped or that unrelated areas change. Apply them one after another by selecting each result in Files.

Say what must stay exact

If a label, sign or face must not change, write it down, for example "keep the text on the bottle exactly as it is". Proofread any lettering in the result anyway; generated text can still come out wrong.

Describe format and style in words

With no size control, the prompt is where you ask for "a vertical portrait format" or "a wide banner". Name the medium as well, such as "flat vector illustration" or "35 mm film photo".

Leave cutouts to a dedicated tool

There is no transparency setting here. For a subject on a transparent background, run the result through Remove Background.

What the AI actually does to your photo

This is not a filter that shifts pixel values. In Edit mode the image model looks at your photo and paints a new image that follows your instruction. That is why it can remove objects, change clothing or restyle a whole scene, and also why areas you never mentioned can drift slightly. Skin texture, hands, lettering and fine patterns are the details to check first.

The result is a new file. Camera metadata from your original, such as EXIF data and GPS location, is not copied into it, and its pixel dimensions and file size can differ from the upload.

Formats and limits

Input formats: JPG, PNG, WebP, HEIC, HEIF, AVIF, BMP and TIFF in Edit mode. Create mode needs no file.
Output format: The format the model returns, normally PNG
Maximum file size: 200 MB per upload
Files per job: 1 source image in Edit mode, none in Create mode. Usually one result; if the model returns more images, the extra ones are added to Files.
Processing: On our servers, which pass the request to OpenAI's image model. A request can run for up to 15 minutes and is not retried automatically.
Free requests and credits: A limited number of free requests per rolling 24 hours, shared between Edit and Create. After that, guests must log in and each request costs credits. The message you see at that point states the current numbers.
Failed requests: Charged credits are refunded automatically. The request still counts toward your free allowance.
Watermark: None added by this site
Storage: Uploads and results are deleted automatically after 48 hours for guests and after 30 days for free accounts. Paid plans can keep files longer.

Frequently asked questions

Why did I get a message instead of an image?

The model replied with text instead of drawing, or the request was declined under OpenAI's content rules. Rephrase the prompt and run it again. Any credits charged for the failed request are refunded, but it still counts as one of your free requests.

Why did parts of my photo change that I didn't mention?

Every edit is a full redraw of the image, not a local retouch. Name the elements that must stay exact and keep to one change per request. Drift can add up over several rounds of re-editing, so for a fresh attempt start again from the original in Files.

Why is the result a different size than my original?

The model picks the output dimensions itself, and the tool does not scale the result back to your original size. Resize or crop the result afterwards if you need exact pixel dimensions.

Is my photo sent to a third party?

Yes. The prompt and, in Edit mode, the image are sent to OpenAI, whose model creates the result. Your upload and the result are stored on our servers for your account's retention period, as with every other tool.

Why am I asked to log in, or told I don't have enough credits?

Your free requests for the last 24 hours are used up. Guests are counted per IP address and signed-in users per account, and Edit and Create share the same allowance. Guests can log in to continue with credits; signed-in users need enough credits for the next request, or can wait until older requests fall out of the 24-hour window.