Image to Prompt: get the prompt from an AI image
Drop an AI image and see the prompt, negative prompt, model, seed, steps and Midjourney parameters stored inside the file, each with a Copy button. Free and private: your image is read on your own device and never uploaded.
This image doesn’t carry its prompt (most platforms strip it). Dreamdrive saves the prompt automatically while you create, so you never lose it again.
Which AI tools keep the prompt inside the image
An image file can carry text alongside the picture: PNG text chunks, EXIF, XMP and IPTC fields. Some AI tools write the prompt and settings there; most do not, and almost every route a picture takes after that (a website, a chat app, a screenshot) throws the text away. What this tool can show depends entirely on where the file came from.
| Where the image came from | What the file usually carries |
|---|---|
| AUTOMATIC1111, Forge and other Stable Diffusion web UIs | The full “parameters” text: prompt, negative prompt, steps, sampler, CFG scale, seed, size and model, in PNG chunks or JPEG EXIF (README). This is what their own PNG Info tab reads. |
| ComfyUI | The prompt graph and the workflow as JSON in the PNG (ComfyUI docs), unless it was saved with metadata switched off. This tool reads the positive and negative prompt, sampler settings, seed, model and size from the graph. |
| Midjourney (PNG downloads from midjourney.com) | In the files we have tested, a description with the prompt, its --parameters and the Job ID. |
| Dreamdrive copies on your computer | The prompt, model and settings, written into each PNG when it is saved. |
| ChatGPT, Gemini, Google Flow, Kling, Runway and most other web platforms | In our tests, no prompt. Some add Content Credentials, which record that an image was made with AI but not the prompt; this tool tells you when they are present. |
| Anything saved from social media, a chat app or a screenshot | Almost always nothing: re-encoding a file removes its text (the ComfyUI docs name this as a common reason metadata is missing). |
How it works, and what it will never do
The file is opened by your browser and read byte by byte by the same parser Dreamdrive uses to import files: PNG tEXt, zTXt and iTXt chunks, JPEG and WebP EXIF (ImageDescription and UserComment), XMP descriptions and IPTC captions. Recognised formats are split into fields: Midjourney descriptions with their Job ID, Stable Diffusion “parameters”, ComfyUI prompt graphs and Dreamdrive’s own files. Anything else that is plain text is shown as it is.
Nothing leaves your device and there are no calls to any service. It never guesses: it does not ask an AI to describe the picture, because a description is a new prompt, not the one that made the image. If the file carries nothing, it says so.
Keep the prompt from the start
Reading a prompt back out of a file only works when the file kept it, and most do not. The reliable way is to keep the prompt at the moment you create. Dreamdrive is a read-only Chrome extension that saves every image and video you make on AI platforms with its prompt, model and settings, in one library you can search by any word you remember. See how it works on Midjourney, ChatGPT, Kling and the other platforms, or read the Learning Centre.
Questions
Is my image uploaded anywhere?
Why does it say my image has no prompt?
Can it read Midjourney prompts?
Does it work like PNG Info in Stable Diffusion?
Will it describe an image that has no prompt?
How do I never lose a prompt again?
Sources
- AUTOMATIC1111 stable-diffusion-webui README: generation parameters are saved with the image, in PNG chunks for PNG and in EXIF for JPEG; PNG Info restores them
- ComfyUI docs: Workflow metadata: PNGs carry `workflow` and `prompt` JSON; metadata is missing with --disable-metadata or after another app re-encodes the file