furybot

Steps

1. Upload the photo you want to keep. This is an edit, so the input is the thing being preserved. Start from the version at the resolution you want out — edits come back at the size you supplied rather than being quietly downscaled.
2. Describe the change, not the picture. "Add a red beanie" works. Re-describing the whole scene with a beanie in it does not — it invites the model to regenerate everything you just described, which is how the face drifts.
3. Name the thing precisely when there is more than one candidate. "Change the jacket" in a photo of three people is a question, not an instruction. "Change the jacket on the man on the left to grey" is an instruction.
4. Add a reference photo to copy a detail. Supply a second photo and ask for the hairstyle, clothing, expression or lighting from it. The feature transfers onto your original rather than producing a new person.
5. Stack edits one at a time. Run one change, download or re-upload the result, then run the next. Three instructions in one prompt is where a model starts negotiating with itself about which to honour.

Tips

• If the identity drifts, your prompt is probably too descriptive rather than too vague. Cut it back to the change itself.
• Removing something is harder than adding something — whatever was behind it has to be invented. Expect more attempts.
• Editing a photo of a real, identifiable person is fine, and the result carries a small AI-generated label for disclosure.

Limits worth knowing before you start

• Anything you do not mention is preserved, but "preserved" is a very high bar rather than a guarantee at the pixel level. Compare at full size if the picture matters.
• It is an edit, not a compositor. It will not give you layers, a mask, or a non-destructive history.
• Nothing is archival. Uploads are deleted as soon as the job finishes, generated files are swept automatically after about a day, and the prompt text is blanked on the same schedule. Download what you want to keep — this is a workshop, not an archive.
• Adult content is absent from this product rather than switched off. There is no setting, plan, administrator action or support request that introduces it, on the site or on the API.
• Four things are refused on every plan, each one traceable to a law rather than to our taste. They are listed in full, with the statute behind each, on the safety page.

Questions

Will it change the whole photo?

No. Only what you ask for. Anything you do not mention is preserved.

Can I edit a photo of myself?

Yes. Edits that depict a real, identifiable person carry a small AI-generated label for disclosure.

Does it shrink my photo?

No. Edits come back at the resolution you supplied.

Account, billing and retention questions are on the FAQ.