PNG Chunk Inspector

A PNG is a fixed signature followed by a sequence of chunks, and nothing else. Each chunk carries a four-byte length, a four-letter type, its payload, and a CRC-32 — computed over the type and the payload but not over the length, a detail that trips up anyone writing a chunk by hand. This inspector walks the whole file in the browser, lists every chunk with its offset and size, pulls the image's dimensions, bit depth and colour type out of IHDR, and decodes tEXt and iTXt blocks, which is where editors leave signatures, copyright lines and camera notes that nothing in the picture shows.

The type code is not just a label — the capitalisation of each of its four letters carries a flag, encoded in a single bit of each byte. The first letter tells a decoder whether the chunk is critical or ancillary: uppercase means the file cannot be displayed correctly without understanding it, lowercase means it can be skipped. The second distinguishes a public, registered type from a private one. The third is reserved and must be uppercase. The fourth is the clever one: it tells an editor that does not recognise the chunk whether copying it into a modified file is safe. A chunk marked unsafe-to-copy depends on the pixels, so preserving it through an edit would produce a file that quietly disagrees with itself.

That scheme is why PNG has aged well. A decoder from 1996 can open a file using chunk types invented since, ignore the ones it does not know, and still render the image correctly — because the file itself tells it which ignorance is safe.

A few structural rules make a file valid rather than merely parseable. IHDR must come first and IEND last. Image data can be split across several IDAT chunks, and they have to be consecutive, because the compressed stream runs across the boundaries as if the splits were not there. A chunk's declared length may not exceed 2³¹ − 1 even though the field holds 32 bits. And the two text chunks differ in a way that matters for anything non-English: tEXt is Latin-1 and cannot represent most of the world's writing, while iTXt is UTF-8 and can, optionally compressed. A file with accented characters mangled in its metadata usually has them in the wrong one.