PNG Info, online.

What is the PNG Info tab, and why does everyone go looking for it?

Automatic1111's web UI has a "PNG Info" tab that does one specific job: you drop in an image it (or a compatible fork) generated, and it reads back the exact prompt, settings and seed baked into that file. It became the standard name for the workflow because so many people needed it for the same reason — they found an image on disk, in a Discord channel or in an old project folder, and had no idea what produced it anymore. This page does the same job, without requiring a local Automatic1111 install: drop a PNG above and get the same information back in a couple of seconds.

ComfyUI and Automatic1111 don't store the same thing.

A PNG saved by Automatic1111, Forge or InvokeAI embeds a single text chunk usually called parameters — a flat block of text with the prompt, negative prompt, sampler, steps, CFG scale, seed and model hash. That's what the classic PNG Info tab was built to parse, and it's a solved, well-documented format.

ComfyUI works differently. It doesn't save a summary — it saves the entire node graph that produced the image, as a JSON workflow chunk, plus a separate prompt chunk describing the resolved API call. That graph can be reopened directly in ComfyUI to reproduce the exact generation, which is powerful, but it also means "reading" a ComfyUI PNG means walking a graph of nodes to find the ones that actually set the prompt, checkpoint and seed — there's no single flat field to grab. This tool resolves that graph for you and shows the same kind of summary you'd expect from an A1111 file, while still exposing the full raw JSON underneath for anyone who wants to open it back up in ComfyUI or inspect a custom node it didn't recognize.

Why this runs entirely in your browser.

The PNG never leaves your device. Reading these text chunks doesn't require a server — it's a parsing problem, not a computation problem — so the whole thing happens in JavaScript, in the tab you're looking at right now. That matters for two practical reasons: prompts sometimes contain material people don't want sitting on a third-party server, and the tool keeps working even if you're offline after the page has loaded once.

When a PNG has no metadata at all.

Sometimes this tool — or the original PNG Info tab — will come back empty, and that's not a bug. The three common causes: the file was re-encoded by an editor or converter that rebuilds the PNG's chunks from scratch and drops anything it doesn't recognize; it passed through a messaging app or social platform that strips metadata on upload for privacy and file-size reasons; or it's a screenshot, which never had generation metadata embedded in the first place because the screenshot tool has no idea a Stable Diffusion or ComfyUI process made the pixels on screen. If the file is a direct, untouched save from the generator, the metadata should still be there.

One image versus a whole folder.

This page is built for the single-file case: someone sent you a PNG, or you found one and want to know what made it. If you're regularly generating dozens or hundreds of images and want to search across all of them by prompt, model, LoRA or seed instead of opening them one at a time, Image MetaHub is the desktop app version of this same parser — it indexes an entire local folder and keeps it searchable, for free.