There is no image-in, image-out model here. The renderer never sees your photograph — it reads words and paints a new picture. That makes some edits easy, some lossy, and two of them impossible. Find out which yours is before you spend anything.
1Your photo, and what you want changed
▣Add your photoDrop it here, or click to choose. It is resized in your browser and re-encoded, which drops the location data before anything is uploaded.
What comes across
What does not
Use one of these instead — they do the real thing, in your browser, free
2The brief
What the writer saw in your photo
Read this first. Everything below was written from it — if it misread your picture, fix the brief before you render.
Your “keep this” clauses, translated
A renderer with no picture cannot keep anything. Each of these had to become a description instead.
What this brief cannot deliver
The prompt that will be rendered
Editable. Everything below re-checks as you type, and nothing here is charged until you press render.
3Render it
This paints a new picture from the words above. It does not edit your photograph, and it never sees it.
Your photoThe render — a new picture, not an edit
4What actually survived
Both pictures go to the writer, which can see them, and it reports which of the brief’s promises the render actually kept.
The one change worth making
Worked examples
Real output from this app, bundled in the page. Open one to see the whole path without signing in or spending anything.
The briefs themselves are recorded model output and are shown in English, the language they were produced in.