Cross-platformWorkflow & organisation
How to Find the Prompt Behind an AI Image You Made Weeks Ago
To find the prompt behind an old AI image, start where you made it: Midjourney's Organize page searches your prompts, Ideogram's Details panel shows the prompt on your images, ChatGPT keeps it in the chat, and AUTOMATIC1111 and ComfyUI write it into the PNG file. If you only have the file from a hosted platform, the prompt is usually not in it. To avoid the search next time, save images with their prompts as you make them, for example with Dreamdrive.
You have the image. You need the prompt: to make a variation, to answer a client, to reuse a look. How hard that is depends entirely on where the image came from. Work through the steps below in order.
Step 1: check the platform you made it on
| Platform | Where the prompt is |
|---|---|
| Midjourney | Organize or Create page: search words from the prompt or attributes; open the image to see the prompt and parameters |
| ChatGPT | In the conversation that made the image; the Images view shows the picture but not the prompt |
| Ideogram | The Details panel on your image shows the prompt and settings |
| Runway | Assets, with a search bar and filters; generations collect under All Generations |
| Krea | The Assets grid of past generations, grouped by batch |
| Leonardo | Your Library and Collections |
| Adobe Firefly | The Generation history panel |
For Midjourney, search for the most unusual word you remember rather than the obvious one: "lighthouse" narrows the results far more than "cinematic". Our Midjourney guide has more search tricks.
Step 2: check the file itself
Some tools write the prompt into the image file:
- AUTOMATIC1111 stores the generation parameters in the PNG. Open the PNG Info tab and drop the image in.
- ComfyUI embeds the prompt and the full workflow. Drag the image back into ComfyUI to load it.
- Hosted platforms such as ChatGPT and Firefly add content credentials that identify the image as AI-made, which is not the same as storing your prompt. Don't count on finding it there.
If the image has been through social media, a screenshot or a format conversion, assume any metadata is gone.
Step 3: if you can't find it
Reconstruct rather than guess. Describe the image as a prompt (subject, composition, light, style, lens), run it on the same platform and model, and iterate. Keep the result with its new prompt this time.
Never search again
The only reliable fix is to save the prompt at the moment the image is made. Dreamdrive, a free Chrome extension, does this automatically on the AI platforms you use: every image and video lands in one library with its prompt, model and settings, searchable by any word you remember, across all your tools at once. Related: how to keep one prompt library across AI tools.