PNG Metadata to JSON
Parse a PNG's binary chunk structure and extract embedded tEXt/iTXt/zTXt text metadata as JSON — nothing is uploaded.
How PNG metadata extraction works
A PNG file is a fixed 8-byte signature followed by a sequence of length-prefixed chunks ([4-byte length][4-byte type][data][4-byte CRC]). Alongside required chunks like IHDR (dimensions, bit depth, color type) and IDAT (pixel data), encoders and editors often embed free-text metadata in three optional chunk types: tEXt (plain Latin-1 key/value pairs, e.g. Author or Description), zTXt (the same, zlib-compressed), and iTXt (UTF-8 text that supports international keywords and can optionally be compressed too). This tool reads the raw bytes of your file with a DataView, walks the chunk list, and decodes any text chunks it finds — using the browser's native DecompressionStream("deflate") API to inflate the zlib-compressed chunks — then serializes everything to JSON. It uses the same low-level binary-parsing approach as the EXIF Viewer, just applied to PNG's chunk format instead of JPEG's marker segments. Common places to find embedded PNG text metadata include screenshots (software/OS info), AI-generated images (prompt/model parameters), and files exported from design tools (author, copyright, or software comments).
Built and maintained by Meet Shah · Last updated
What this tool is used for
- Reading the text chunks a PNG carries, which most viewers never show.
- Checking whether a generated image embedded prompt or tool metadata.
- Inspecting a PNG's chunk structure to understand a file.
- Confirming an export stripped metadata as intended.
- Finding a colour profile or gamma chunk.
Frequently Asked Questions
- What metadata does a PNG carry?
- Text chunks — `tEXt`, `zTXt` and `iTXt` for compressed and international text — plus `pHYs` for physical dimensions, `gAMA` and `iCCP` for colour, and `tIME` for modification time. All of them are optional and most tools write only a few.
- Where does AI image metadata live?
- In a `tEXt` chunk, usually keyed `parameters` or `prompt`. Stable Diffusion and several other tools write the full generation parameters there, which is why an image can carry its own prompt, seed and model — and why sharing one discloses all of it.
- How is PNG metadata different from EXIF?
- PNG uses its own chunk structure rather than EXIF, though an `eXIf` chunk was added later to carry it. That is why a camera JPEG converted to PNG typically loses its EXIF entirely — the container changed and most converters do not migrate it.
- Does a canvas round trip preserve metadata?
- No, and this matters. Re-encoding through a canvas produces a fresh PNG with no text chunks at all, so any edit in a browser silently strips everything. That is a privacy improvement and a real loss if the metadata was the point.
- How does chunk structure make this readable?
- Every chunk carries a length, a four-character type, its data and a CRC — so a parser can walk the file and skip anything it does not understand. That design is why PNG has been extended repeatedly without breaking older readers.
Common errors and gotchas
- Assuming a PNG carries EXIF the way a JPEG does, when its metadata lives in different chunks.
- Publishing a PNG whose text chunks still hold prompts, paths or tool details.
- Expecting a viewer to show these chunks, since most ignore them entirely.
- Confusing tEXt, iTXt and zTXt, which differ in encoding and compression.
- Treating an absence of text chunks as an absence of identifying information.