filehugger.comHow file formats actually work
Long-form writing about the insides of the formats this site converts — the bytes, the specifications and the measurements, by the engineer who wrote the converters. Every claim is either sourced to a specification or measured on a real file, and every article says what its method cannot tell you.
What is actually inside a PDF: writing one by hand, 588 bytes at a time
A PDF is not a picture of a page. Here is a complete, working PDF typed out byte by byte — the objects, the cross-reference table, and why the file is read from the bottom up.
PNG has five filters, and the one the spec recommends costs this site 23%
The five PNG filter types measured on real files, with the encoder reproduced byte for byte. On interface screenshots the standard adaptive heuristic loses to doing nothing at all — and here is the measurement that explains why.
JPEG forgets in eight-by-eight squares, and the arithmetic is on the page
What a JPEG actually discards, measured: the real quantisation tables an encoder writes, one 8x8 block taken apart coefficient by coefficient, and the block grid growing as a number you can watch.
A HEIC photo is nine pictures and an 8-byte note on how to staple them together
Why HEIC is so much smaller than JPEG and why so little software opens it, both measured: the real size-versus-quality curve, the point where HEIC stops improving, and the item structure a decoder has to walk before it finds a single pixel.
The .webp extension hides two unrelated formats, and only one of them can be perfect
A lossy WebP and a lossless WebP share a container and an extension and nothing else. Measured: where each one wins, the quality ceiling lossy WebP can never pass, and why a transparent lossy WebP contains two codecs at once.
The first article is not out yet. It will be here shortly.