Free. Read-only. Nothing uploaded

ComfyUI Workflow Extractor

Drop a ComfyUI image to get its workflow.json, prompt, seed, models and LoRAs, and to see whether it exposes your computer username or local folder paths.

Drop an AI image, or click to choose

PNG, JPEG, WebP. Read in your browser, nothing is uploaded, up to 50 MB

Where ComfyUI keeps the workflow

ComfyUI's Save Image node writes two JSON text chunks into every PNG: workflow, the editor graph you see on the canvas, and prompt, the API-format graph that actually executed. ComfyUI's WebP save node puts the same JSON into EXIF text tags instead. This tool reads PNG and WebP.

Flux, SDXL, SD 1.5, video and upscale workflows are all stored the same way, because the format belongs to ComfyUI, not to the model.

Check what your images give away

A workflow is a complete recipe: prompts, seed, sampler settings, and the exact checkpoint, LoRA and VAE files. Some custom nodes (path-based loaders, some save nodes) also store absolute paths like /Users/jordanlee/ComfyUI/models/..., which reveal the account name on the computer that made the image. When this tool finds a username or a local path, it says so at the top of the results.

To share an image without any of it, run it through the ComfyUI metadata remover. Keep your original or the downloaded workflow.json if you want to reuse the graph.

Loading the workflow back into ComfyUI

  1. Download workflow.json above, or keep the original PNG.
  2. Drag either file onto the ComfyUI canvas.
  3. Install any missing custom nodes ComfyUI reports, and the models named in the loader nodes.

Frequently asked questions

How do I get the workflow from a ComfyUI image?

Drop the original PNG (or a WebP saved by ComfyUI) above. If the workflow is still inside, you can view it and download it as workflow.json, then drag that file onto the ComfyUI canvas. You can also drag the original PNG onto the canvas directly.

Why is there no workflow in my image?

The file was probably re-encoded or exported. Screenshots, image editors, format conversions and many upload pipelines drop the text chunks that hold the workflow. The image may also have been saved with ComfyUI's --disable-metadata flag, or stripped on purpose by its creator.

What is the difference between the workflow and the API prompt?

ComfyUI saves two things. The workflow is the editor layout with every node, position and link, which is what drag-and-drop restores. The prompt is the API-format graph of what actually ran. Some files only have the prompt; it still shows the settings, but it will not rebuild the editor view the same way.

Will the workflow run on my machine?

Only if you have the same custom nodes and models installed. ComfyUI will list missing node types when you load it. Model and LoRA file names are in the workflow, so you can see what to download.

Is the image uploaded?

No. The file is read in your browser tab and never sent anywhere, which you can confirm in your browser's network panel.

Related