To edit a photo in AI Media Studio, upload your image, describe the change and what to keep, generate a result, then review and download it. Start with the AI image editor, not a template library. You need to sign in to generate an edit.
What do you need before editing?
Use a photo you have permission to edit and publish. Keep the original file. A clear JPG, PNG or WebP is a practical starting point; the upload limit is 10 MiB. If a phone photo will not open, export a JPG and try again. Do not upload confidential documents or someone else's photo without permission.
Every edit needs a paid plan. Plans start at $19/month. Check the generation cost before continuing. Commercial licensing on paid plans does not remove your responsibility for source-image rights.
Step 1: upload the photo you want to change
Open the editor and choose your photo, or drag it into the upload area. Wait until it appears before generating. For this walkthrough, use a portrait with a distracting background. You can replace the source photo before starting; do not expect to change it while a generation is running.
The example below is an instruction to try, not a documented before-and-after result or a promise of perfect preservation.
Step 2: describe the edit and what should stay the same
Write in your own language. Name the change, the subject and any constraints. For the portrait example:
Replace the busy background with a plain warm-grey wall and soft daylight. Keep the person's face, hair, clothing and pose unchanged. Keep the original framing. Do not add text or objects.
This is more useful than “make it better” because it defines both the new background and the details to preserve. Start with one main change rather than a list of unrelated transformations. A preservation instruction helps guide the model, but cannot guarantee an unchanged face or every original pixel.
Step 3: generate, compare and check the details
Sign in if prompted, check that your workspace has a paid plan, and generate the image. Wait for the result; completion time depends on the model and service conditions, so there is no fixed 30-second promise.
Use Compare to inspect the original and result. Check the face, hair edges, hands, clothing, background and shadows. For product photos, also check shape, labels and colour. Do not accept a result just because its background looks attractive.
If a detail is wrong, adjust the instruction or use Refine to continue from the selected result. For example: “Keep this background, but restore the original jacket colour.” Refinement is another generation and uses credits from your paid plan. Return to the original if successive edits drift too far.
Step 4: download the version you actually reviewed
Select the result you want and choose Download. The editor saves that image directly; this is not a multi-format export wizard. Open the downloaded file to confirm it is the intended version and that its dimensions suit your destination.
For a website, prepare the delivered file using the web image resizing guide. Enlarging or compressing a file is a separate task from changing its content.
What should you do if the result is not usable?
Simplify the request, name the problem and specify the important details to retain. Use a clearer source photo when the original subject is blurred or obscured. AI can invent fine detail, distort text or change identity; it cannot certify that an edited photo is an accurate record. Use a manual editor when exact pixel-level control is required.
For the next edit, try a single, concrete task such as removing a background distraction or changing the visual style. Review every result before sharing or publishing it.