AI image detector by generator

ComfyUI image detector: the workflow saved in the PNG

ComfyUI writes its API prompt graph and editable workflow into the PNG as text chunks unless it is started with --disable-metadata, and isGenAI reads that graph to show the saved prompt, negative prompt, checkpoint and sampler settings.

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.

What ComfyUI writes

ComfyUI’s documentation describes two records saved with generated images: the prompt field, an API execution graph of node classes and inputs, and the workflow field, the editable graph with layout. Dragging a generated PNG back onto the canvas restores the full workflow including seeds. The --disable-metadata command-line option turns this off, with the help text “Disable saving prompt metadata in files”.

Both records are plain JSON in PNG tEXt chunks. They are not signed, and any tool can edit them.

How isGenAI reads the graph

The checker parses the prompt graph, starts from the SaveImage or PreviewImage node, follows the image path back to the sampler and reads the positive and negative CLIPTextEncode text, the checkpoint name and the sampler settings. With exactly one sampler it reports a saved generation prompt; with several samplers or stages it reports a saved workflow, because a multi-stage graph has no single original prompt.

The ComfyUI workflow viewer shows the complete node inventory and lets you download the JSON unchanged.

What a repost removes

Any re-encode drops the text chunks: screenshots, format conversion, social uploads and most chat apps. WebP or JPEG saved by other nodes may hold the record elsewhere or not at all. Keep the original PNG from the output folder.

A saved graph proves what the software wrote; it does not prove who ran it. isGenAI reports it as a declaration of moderate strength.

How to check an image from this generator

  1. Get the original download, not a screenshot or a copy saved from a chat or social feed.
  2. Upload it to the free AI image detector. Files up to 20 MiB in PNG, JPEG, WebP, AVIF or TIFF are read on the server and are private by default.
  3. Open “How we checked” and read each finding with its source field: signed record, declared source type, saved prompt, AI statement or matching copy.
  4. Treat a clean result as “no evidence found”, never as proof that a person made the picture.

Open the ComfyUI workflow viewer

Frequently asked questions

Does ComfyUI add metadata to images?

Yes by default. It saves the API prompt graph and the editable workflow as PNG text chunks, and the --disable-metadata option turns that off, per ComfyUI’s documentation and command-line help.

Why does isGenAI show a workflow but no prompt?

The graph has more than one sampler or stage, or the text encoders disagree. isGenAI reports a saved workflow rather than guessing which prompt produced the final image.

Is a ComfyUI image detectable after a screenshot?

No. The text chunks are gone and isGenAI does not infer origin from pixels. Use the original PNG.

Can I get the workflow JSON back?

Yes. The ComfyUI workflow viewer displays the saved workflow and prompt records and lets you download the JSON exactly as stored in the PNG.

Sources

Vendor behaviour changes. These pages describe what the vendor documents and what isGenAI reads from a file; they do not test every release.