Home / Get the ComfyUI workflow from an image

Get the ComfyUI workflow from an image

Drop a ComfyUI PNG and download the complete workflow as JSON – even without ComfyUI on this device.

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. Take the PNG straight from ComfyUI's output folder. A screenshot, or an image that went through a messenger, no longer contains the workflow.
  2. Drop the file here: drag it on a computer, or tap the drop area on a phone.
  3. Check the heading: if it says “Created with ComfyUI”, the data was found. Below you'll see the prompt, models, LoRAs and seed.
  4. Tap “Download workflow (.json)”. The file is named after your image with the ending -workflow.json and is neatly indented.
  5. Drag the JSON file onto the ComfyUI canvas or open it from the menu there. The node graph appears as it was when the image was generated.
  6. If ComfyUI then shows red nodes, check the JSON file with the “ComfyUI workflow checker” tool. It lists which custom nodes and models are missing.

What you can use it for

  • Save a workflow on the go: You only have the image on your phone and no access to your computer. You can still pull out the workflow as a small JSON file and send it to yourself.
  • Share a workflow cleanly: Instead of sending a large PNG that may be recompressed on the way, share the JSON file. The node graph is guaranteed to stay intact.
  • Keep versions in order: If you tweak a workflow a lot, you can store the matching state as JSON next to every good image and return to exactly that version later.
  • Trace older results: For an older image you can see which sampler, steps and models were used back then, even if you changed the workflow long ago.

Where the workflow lives

ComfyUI stores two text fields in every PNG it creates: “prompt” with the settings and “workflow” with the complete node graph. Drag the PNG into ComfyUI and you get the workflow back – but only with ComfyUI.

View it without ComfyUI

This tool reads the fields directly in your browser. You instantly see the prompt, model, LoRAs, sampler and seed, and can save the workflow as a JSON file, for example to share it or load it later.

If the image was sent via Discord, Reddit or social media, the fields are often missing. Then only the original file helps.

Go to tool: Read an AI image: prompt and workflow

Frequently asked questions

Does it work with Flux workflows?

Yes. Workflows with SamplerCustomAdvanced, FluxGuidance and UNET loaders are recognized too.

Are custom nodes missing from the JSON?

The workflow contains all nodes. Whether they run for you depends on whether you have the custom nodes installed.

Why do I only get “Download API prompt”?

Then the PNG only contains the “prompt” field, not “workflow”. That file holds all nodes and their values, but not their layout on the canvas.

Does this work with JPG or WebP from ComfyUI?

The download button is available for PNG files in which ComfyUI stored the workflow as a text field. For other formats the tool shows the prompt and settings if they are in the file.

Is the downloaded file identical to the workflow in the image?

In content, yes. It is just saved with indentation so you can read it comfortably in a text editor.

I have several images from the same session. Which one should I use?

Each PNG contains the workflow as it was when that particular image was created. Use the image whose result you want to repeat.

Are prompts from ComfyUI images saved?

The prompt found is added to your prompt collection, which lives only in this browser. You can delete it there.