Image
Image to Prompt
Upload a picture. You get three prompt styles in this page’s language, and a link that opens the thumbnail tool with the prompt filled in.
Upload a picture, then write a prompt from what is in it.
What image to prompt writes
Upload a picture and this page writes a prompt that describes it. You get three versions in the language of this page: plain text, a Midjourney-style line with aspect ratio and stylize, and a Flux or Stable Diffusion line made of short phrases. Copy the one you need, or switch formats without uploading again.
The search behind the page is practical: “image to prompt,” “turn a photo into a Midjourney prompt,” and “Flux prompt from a reference image.” The text follows what is visible. It should not invent a caption that is not in the picture, and it does not name a private person.
Generate with this prompt opens the thumbnail tool on this site with the current text filled in, so you can try the wording as an image without copying it by hand. The uploaded file is resized in the browser and is not stored here.
How to get a prompt from a picture
Upload the reference
Use JPG, PNG, or WebP. A picture with one clear subject produces a tighter prompt than a collage. The browser shrinks the long side before sending it.
Read the plain version first
Plain text is the full description: subject, place, light, framing, and color. If a detail is wrong, that is the version to judge. The other two are the same idea in another syntax.
Switch to Midjourney or Flux
Midjourney ends with aspect and stylize flags and does not add a version flag. Flux is comma-separated phrases without those flags. The format menu changes the text already on the page.
Copy it, or generate here
Copy puts the visible prompt on the clipboard. Generate with this prompt opens the thumbnail tool in this language with that text in the topic field, trimmed if it is very long.
Prompts people extract from a picture
A product photo into a reusable prompt
Upload a bottle on a wood table. Plain text should mention the bottle, the table, and the light. Flux should be short phrases of the same scene. That is the usual “describe this product photo as a prompt” job.
A landscape for Midjourney
A wide coast at dusk becomes a Midjourney line with a wide aspect flag. Use that version when the next tool is Midjourney. Use plain text when you are still editing the idea.
Try the wording on this site
After the prompt appears, use Generate with this prompt. The thumbnail page opens with the text filled in, so you can see whether the description is specific enough to make another image.
Tips for a prompt that matches the picture
- Crop to the subject you care about before uploading. Extra background becomes extra prompt.
- Judge the plain text before you trust the flagged versions.
- If readable words are in the image, check that the prompt quotes them instead of inventing new ones.
- Private names should not appear. The description is of the picture, not of an identity.
- The generate link uses the format currently selected, so switch first if you want the Flux line carried over.
Image to prompt questions
Which prompt formats do I get?
Plain text, a Midjourney-style prompt, and a Flux or Stable Diffusion prompt. All three come from one upload. The menu switches among them.
What language is the prompt in?
The language of this page. A French page writes the prompts in French. Switch the page if you want another language, then upload again.
Can I generate an image from the prompt here?
Yes. Generate with this prompt opens the YouTube thumbnail tool in this language with the current text filled in. You can still copy the prompt and use it somewhere else.
Does the tool name people in the photo?
It is instructed not to name a private person. It describes what is visible: clothing, place, light, and composition.
Is the upload stored?
No. The browser resizes the image and it is sent only to write the prompt. This site does not keep the file.
Why are there Midjourney flags in one version only?
Those flags belong to that syntax. The Flux version stays as phrases, and the plain version stays as sentences, so you can paste each one into the tool that expects it.