png::meta tEXt · iTXt · zTXt

ComfyUI Workflow Extractor

Drop a ComfyUI-generated PNG to pull out the complete workflow JSON, the executed prompt graph, every CLIPTextEncode prompt, and the seed / sampler / model that produced it.

🎥 [Demo: Upload a PNG to see AI prompts & workflow data extracted in seconds]

Drop a PNG here or click to browse · JPEG & WebP work too
ComfyUI · A1111 · Forge · InvokeAI · SwarmUI · Fooocus · any tEXt/iTXt/zTXt · JPEG/WebP EXIF

What ComfyUI stores in a PNG

When ComfyUI saves an image with the default Save Image node it writes two PNG text chunks: workflow — the editable canvas graph (nodes, positions, links, groups, widget values) — and prompt — the flattened API graph that was actually executed (node id → class_type + inputs). Both are plain JSON, so the entire recipe travels with the image. Some custom save nodes also add parameters in A1111 style for Civitai compatibility.

How to extract and reload the workflow

  1. Upload the original PNG above (not a screenshot — see why PNGs lose metadata).
  2. The Generation summary shows seed, steps, CFG, sampler, scheduler, checkpoint, LoRAs and latent size, plus every text prompt found in CLIPTextEncode nodes.
  3. Click download workflow .json, then drag that file onto the ComfyUI canvas — or drag the PNG itself into ComfyUI. Missing custom nodes will show as red placeholders; install them via ComfyUI-Manager.

prompt vs. workflow: which one do you need?

Use workflow to rebuild the graph in the UI. Use prompt if you are calling ComfyUI's HTTP API (POST /prompt) from a script — it is exactly the payload the server expects. PNGMeta pretty-prints both and lets you copy or download each separately.

Common problems

  • Only "prompt" is present, no "workflow" — the image was produced via the API or with a save node that skips the canvas graph. You can still see all prompts and settings.
  • No metadata at all — the file was re-encoded (Discord mobile, WhatsApp, Reddit, screenshots). Ask for the original file.
  • Prompt text is empty — some workflows feed prompts through custom text nodes (e.g. Primitive, String Literal); open the raw workflow chunk and search for the node title.

Frequently asked questions

How do I get the workflow from a ComfyUI image?

Upload the PNG to pngmeta.app, then click "download workflow .json" in the generation summary. Drag the JSON (or the PNG itself) onto the ComfyUI canvas to restore the graph.

Why does the ComfyUI PNG only contain "prompt" and not "workflow"?

Images generated through the ComfyUI API or with certain custom save nodes only embed the executed prompt graph. The prompts and settings are still visible; the canvas layout is not.

Can I extract a ComfyUI workflow from a JPEG or WebP?

Only if the tool that saved it wrote the workflow into EXIF/XMP. PNGMeta reads EXIF UserComment and XMP from JPEG and WebP, but most ComfyUI save nodes only embed metadata in PNG.

Related guides