Home / Read an AI image: prompt and workflow

Read an AI image: prompt and workflow

Shows which prompt, model and settings were used to create an AI image. You can download ComfyUI workflows directly.

Drop an AI image here

or tap to choose. Ideally the original file (PNG), not one from social media.

Choose file

Step by step

  1. Drop the original file as PNG, JPG or WebP, exactly as it came out of the image generator. One file is analyzed at a time.
  2. Read the heading: it names the program that created the image, for example ComfyUI, Automatic1111 / Forge or NovelAI.
  3. Copy the prompt and negative prompt with the “Copy” button below each text box.
  4. Check model, LoRAs, seed, steps, CFG, sampler and scheduler under “Settings”. For ComfyUI images, guidance, denoise and image size may be listed too, depending on the workflow.
  5. For ComfyUI images, save the node graph with “Download workflow (.json)”. If the workflow is missing but the prompt is there, you get “Download API prompt (.json)” instead.
  6. Finally, look at “Other data in the file”. It lists camera, software or author, and a warning appears if a GPS location is stored.

What you can use it for

  • Recreate your own images: You made a great image weeks ago but no longer remember the settings. With the seed, sampler and model from the file you can get very close to it or build deliberate variations.
  • Understand images from galleries: Some platforms and forums offer the original file for download. Then you can see which models and LoRAs are behind an image you like.
  • Build a prompt archive: Every prompt you read is added automatically to your prompt collection below the result. You can search it, delete entries and export everything as JSON.
  • Check before sharing: Before you pass an image on, see what the file reveals: prompt, workflow, camera data or a location. Then decide whether to remove that data first.
  • Gauge the origin: If a file carries Content Credentials (C2PA), the tool shows which software is named and whether, according to that data, AI was involved.

Which images work?

  • ComfyUI: prompt, model, LoRAs, seed, sampler and the complete workflow.
  • Automatic1111 and Forge: prompt, negative prompt and all settings, also from JPG and WebP.
  • NovelAI: prompt and settings.
  • ChatGPT, Adobe Firefly and others: a note on Content Credentials (C2PA), if present.

Instagram, WhatsApp, Facebook and most other platforms strip this data on upload. After that, nothing is left in the file.

Related guides

Frequently asked questions

How do I get the ComfyUI workflow from an image?

Drop the image here and choose “Download workflow”. You can drag the JSON file straight into ComfyUI.

Why is nothing shown for my image?

The file no longer contains generation data. That happens when an image was sent via social media or messengers. The original file usually works.

Can the tool tell whether an image is AI-generated?

Only if the file contains such data, for example a prompt or Content Credentials. Without it, no reliable statement is possible.

Is my image uploaded?

No. The file is only read in your browser.

Where is my prompt collection stored?

Only in your browser on this device. In a private window or after clearing browser data it may be gone, so back up important prompts with “Export (.json)”.

What does the number in brackets after a LoRA mean?

It is the strength the LoRA was used with in the workflow. If there is no number, the strength was not stored as a fixed value in the file.

What does “ComfyUI (workflow only)” mean?

The file only contains the node graph, not the executed settings. The texts are then read from the node fields. In that case it can only be guessed which text is positive and which is negative.

What is the difference between workflow and API prompt?

The workflow holds the complete node graph including layout, as you see it in ComfyUI. The API prompt only holds the nodes and their values, without layout.

Why do I see “All text fields in the workflow”?

If a ComfyUI workflow has more than two text fields, for example for several prompts or regions, you'll find all of them in this expandable section.

Can I read several images at once?

No, one file is read at a time. Drop the images one after another; their prompts collect in your prompt collection.