AI Image Metadata Reader (Stable Diffusion, ComfyUI, Midjourney)
Drop a PNG, JPEG or WebP created with Stable Diffusion (Automatic1111, Forge), ComfyUI, InvokeAI, NovelAI or Midjourney and read the generation data embedded in its metadata: positive and negative prompt, seed, steps, sampler, CFG scale, size and model. The file is parsed directly in your browser, including compressed PNG text chunks, EXIF and XMP, and every raw metadata field is available to inspect.
Reviewed by Vincenzo Manto, Founder & Lead Developer · Last updated September 30, 2026
How it works
- 1
Drag an AI-generated image into the drop area, or choose a file.
- 2
The tool reads PNG text chunks, EXIF and XMP metadata locally.
- 3
Review the detected prompt and parameters, and copy them or expand the raw fields.
Screenshots
Example
a cat, masterpiece
Negative prompt: blurry
Steps: 25, Sampler: Euler a, CFG scale: 7, Seed: 12345, Size: 512x768Generator: Automatic1111 / Forge
Prompt: a cat, masterpiece
Negative prompt: blurry
Steps: 25 Sampler: Euler a CFG scale: 7
Seed: 12345 Size: 512x768Where AI generation data lives
Stable Diffusion front ends typically store a text block called parameters inside a PNG, while ComfyUI stores the whole node graph as JSON so a workflow can be restored by dragging the image back in. JPEG and WebP files use EXIF UserComment or XMP fields instead. Reading these fields is enough to reconstruct the prompt, seed and settings that produced an image.
Reproducing and learning from images
With the seed, steps, sampler, CFG scale and model, you can recreate an image or generate controlled variations by changing one parameter at a time. Studying the prompts behind images you admire is also a fast way to learn prompt structure. Respect the terms and licences of the images you inspect.
Frequently asked questions
Why does my image show no prompt?
Metadata only exists if the generator wrote it and nobody removed it afterwards. Social networks, chat apps and many image hosts strip metadata on upload, and screenshots or re-saved copies do not contain it. Use the original file from your generator.
Which generators are recognised?
Automatic1111 and Forge (the parameters text), ComfyUI (the prompt graph), InvokeAI, NovelAI and Midjourney descriptions. Other tools that write similar text are usually shown in the raw fields section.
Does the image get uploaded?
No. The file is read by your browser and never sent to a server, so you can safely inspect private or unreleased work.
Can it read the seed of a Midjourney image?
Midjourney embeds the prompt text and the job ID, but not a reproducible seed in the file. The job ID lets you look up the original job in your Midjourney account.