FAQ

The questions worth asking first.

Including the one where the answer is no.

Does this recover the original prompt?

No, and neither does anything else. The prompt that made an image is not stored in the file. Unprompter reads the picture and writes instructions that land in the same place — which on style, palette, lighting and composition gets you very close, and on a specific face or a brand's exact typeface does not. It tells you what it saw rather than pretending to know what was typed.

What does it cost to run?

Pro is $15 once. The model is billed by your provider directly to you: a 1024px image on a small cloud model is a fraction of a penny, and on a model running on your own machine it is nothing at all. Unprompter never sits in the middle of that transaction, so there is nothing for us to mark up.

Why bring my own API key instead of buying credits?

Because credits mean a server in the middle: one that holds your images long enough to process them, and adds a margin to every token. Your own key means wholesale pricing, no queue, and no copy of your pictures anywhere. It costs you one paste on the settings page.

Do my images get uploaded to you?

Never. There is no Unprompter server in the analysis path at all — the extension calls your provider directly from your browser. The only thing that reaches us is your licence key when you activate Pro.

What about EXIF and GPS data in my photos?

Stripped before anything is sent. Every image is redrawn through a canvas and re-encoded, which cannot carry EXIF forward. It is not an option you can leave switched off — it is how the resize works, and the build's own test proves it on a real photograph.

Does it work without an internet connection?

With a local model, completely. The palette, the aspect ratio, all nine format renderings, the variations and your whole library are computed in the browser. Only the reading itself needs a model, and that model can be on your own machine.

Which image generators does it write for?

Midjourney, Stable Diffusion and SDXL, Flux, DALL·E and GPT Image, Ideogram, Gemini and Nano Banana, video models like Sora, Veo and Runway, plus plain English and raw JSON for your own pipeline. Nine formats from one reading, and switching between them costs nothing.

Why three browsers?

Because that is how people actually work: a desktop, a laptop, and something else. Seats are tracked on our server, so moving to a fourth machine is a click on the manage page to free one. Reinstalling in the same browser does count as a new seat, since the device id is regenerated.

Is the free tier a trial?

No, it is a tier. 10 images a day, plain-English prompts, every input method, the measured palette and every provider including local models. It does not expire and it does not ask for a card.

What happens if my licence server is unreachable?

Pro keeps working. The extension re-checks weekly and tolerates a fortnight offline before it lapses back to the free tier, so a flight or an outage does not cost you anything. A key we have actually revoked is a different matter: that answer arrives as a refusal rather than a network failure, and takes effect on the next check.

Can I use the prompts commercially?

They are yours. Unprompter claims nothing over what it writes. Whether the image you read was yours to work from is a question about that image, not about this tool, and it is worth asking before you rebuild someone's photograph.

Will it read images of people?

It describes appearance, clothing and lighting, and it is instructed never to guess at who someone is. If you want the look without the person, use style-only mode — it reads medium, light, grade and finish and leaves the subject blank.

Does it work in Edge, Brave or Arc?

Yes — anything Chromium-based on version 116 or newer, which is where the side panel arrived. Each browser uses one of your three seats.

Can I get a refund?

Within 30 days, yes, no argument. Email us and the key is revoked; the extension drops back to the free tier on its next check.