How do I turn an image into a prompt?

Upload the image to the image-to-prompt tool, keep or edit the instruction, pick a model and run it. The prompt appears as text that you can copy.

Does it give me the original prompt?

No. It works from the pixels and writes the prompt that most likely describes what it sees. Treat it as a strong starting point, not the text that was used to make the image.

Does it work on photos and not only AI images?

Yes. Any image can be described this way. For a photo, the result reads as a description of the scene and its look, which you can use to generate something similar.

Which image model should I paste the prompt into?

Any text-to-image model accepts it. Models differ in how literally they follow long prompts, so run the same prompt on two or three and compare the results.

How much does it cost?

It is billed by the tokens the model uses, so a run with one image and the default instruction is small. The tool header shows the credit rate before you start.

Can I get a shorter prompt, or a prompt in another language?

Yes. Replace the default instruction with what you need, such as "one sentence in German" or "keywords only, comma separated".

Can I use the prompts commercially?

A prompt is text you generated from an image. Make sure you have the rights to the image you upload, and check our Terms and the policies of the model provider before you publish what you generate from it.

faktry

Image to prompt

Turn an imageinto a prompt.

Upload any image and get back a prompt you can paste into a text-to-image model.

Create your account Already a member? Open the tool

To turn an image into a prompt, upload it to faktry's image-to-prompt tool and run it. A vision model looks at the picture and writes one dense paragraph covering the subject, composition, art style, lighting, colour palette and mood, ready to paste into a text-to-image model. It is billed by the tokens the model uses, and the rate shows in the tool before you start.

Upload, run, copy.

The prompt comes back as text, so you can edit it before you use it.

01
Upload the image
Any image works: an AI-generated picture, a photo, a screenshot, a piece of reference art. One image per run.
02
Keep the default instruction or change it
The default asks for the likely generation prompt in one paragraph. You can replace it, for example to ask for a shorter prompt, only the art style, or a list of keywords.
03
Pick a model and run
Claude Sonnet 5.5 is the default. Claude Sonnet 5 and GPT-5.6 Terra are also available. The tool shows the credit rate before you start.
04
Copy the prompt and use it
Paste it into an image model, change the parts you want to differ, and generate. Running the prompt on a different model is a quick way to see how each one reads the same description.

What you control.

The result is only text, so the settings are about what kind of text you get.

Instruction: Up to 2000 characters. Ask for what you need: a full prompt, a style-only description, a shorter version, or a different language.
Model: Claude Sonnet 5.5, Claude Sonnet 5 or GPT-5.6 Terra. They describe the same image a little differently, so try a second one if the first misses something.
Length: The default allows about 500 tokens of output, which fits a dense paragraph. You can raise it to 4096 for a longer description.
Temperature: GPT-5.6 Terra only, from 0 to 2. Lower values keep the wording closer from run to run.
Price: Billed by the tokens the model reads and writes, not a flat fee. A single image with the default instruction is a small run.

Models to use the prompt with.

Each has its own page with examples and settings.

Want to change the image instead of describing it?

Questions about turning images into prompts.

Still have questions?

Upload the image to the image-to-prompt tool, keep or edit the instruction, pick a model and run it. The prompt appears as text that you can copy.

Write your first prompt from an image.

Upload an image and see the prompt it describes.

Create your account

The credit rate shows before every run.