A working example, with the files

ComfyUI workflow JSON and API prompt JSON

Choose the record that matches your next task. The editable workflow describes the canvas; the API graph describes node classes and inputs. The field named prompt is JSON and is not necessarily a sentence you can paste into a text box.

Controlled example PNG file that carries this record: ComfyUI workflow found

workflow-records.pngPNG file · controlled example

  1. Signature + IHDR
  2. iTXt XML:com.adobe.xmp
  3. tEXt parameters
  4. tEXt prompt / workflowrecord
  5. tEXt invokeai_metadata
  6. eXIf
  7. caBX C2PA manifest
  8. IDAT pixels
  9. IEND

ComfyUI workflow found

The node graph travelled with the image.

The prompt graph and workflow chunks are listed node by node, so you can see how the image was built.

Open the workflow viewer →
ComfyUI workflow found: where this record sits inside a PNG file and what the isGenAI report says about it. The picture is a code-drawn controlled example that carries this record; download it and check it yourself.

Follow the steps

  1. Download or inspect the two JSON records linked below.
  2. In the editable workflow, find the nodes and links arrays. Follow the connection from node 1 to node 2.
  3. In the API graph, find entries 1 and 2. The SaveImage inputs refer to the LoadImage output.

Read the controlled pair

Both example records describe the same small two-node structure. The editor version additionally records layout values. The API representation uses node IDs as object keys.

The viewer reports the same two node types for both. That agreement is useful when checking the fixture, but it does not prove that an arbitrary pair of graphs has equivalent behavior.

Pick the correct download

Use the editable record when your next task is restoring a canvas in a compatible ComfyUI environment. Use the execution-oriented record when building an integration that expects that shape. Check the requirements of the receiving tool.

The files contain references, not a complete environment. A successful JSON download cannot establish that model files, node packages and input assets are available.

Why there is no text prompt in this example

The example moves an image from LoadImage to SaveImage. It does not include a text-conditioning sampler path. A reader should preserve the available graph rather than invent a positive prompt from the visible pixels.

Download · Run · Compare

Try the verified example

We drew the picture in code and attached controlled records, so you know exactly what each file contains. The saved prompts and model names are demonstration data. These files test record reading end to end; no image generator, screenshot app or social network touched them.

Download editable workflow JSON · Download API graph JSON

Observed output

[
  {
    "field": "workflow",
    "nodes": 2,
    "node_types": [
      "LoadImage",
      "SaveImage"
    ]
  },
  {
    "field": "prompt",
    "nodes": 2,
    "node_types": [
      "LoadImage",
      "SaveImage"
    ]
  }
]

Hashes and reproduction

  • workflow-records.png · 14,013 bytes427aa019b768aaaeb2fb59724a4e49772e041ee6091f47cffc7ec6a35867dc08
  • without-records.png · 13,179 bytesacdc3a0a5e011cf57af9bf0f1da74c75c2d280a2cc9af9297410c529f86932cf

Fixtures and observations are CC0-1.0. Download all observations and versions or read the procedure. The complete Node.js reproduction script (requires sharp) creates the same test inputs and compares bytes and pixels. The isGenAI source checkout additionally verifies saved fields with npx tsx scripts/build-guide-lab.ts.

For the separate export study, explore the 18 metadata-retention cases.

Sources and method

The linked primary sources describe the formats. Every downloadable fixture output was verified with automated assertions against the current isGenAI reader. These guides were prepared with AI assistance and reviewed against those examples.

Report an error or unsupported example. For a signed record, inspect Content Credentials.