Tracking the evolution of wasmpeg against FFmpeg's own FATE regression suite, plus binary size over time.
Does a file decode without erroring, against FFmpeg's own FATE sample suite.
Stricter: does every decoded frame match FFmpeg's own reference output byte-for-byte. This is the number that matters — coverage says we attempt a codec, correctness says we decode it right.
dist/gpl-cpu.wasm — already built for the FATE run above, so measuring it costs
nothing extra. It's ~9 KB bigger than the LGPL cpu.wasm package (two extra
encoders, everything else identical), close enough to track as one number. For
scale: @ffmpeg/core is currently ~61.7 MB unpacked — wasmpeg trades codec breadth
for size, on purpose, not by accident.
Refreshed daily by .github/workflows/track.yml: builds
wasmpeg's gpl-cpu target, runs compat.mjs/fate.mjs against the FATE suite, and
size.sh's measurement, then appends one point per metric to the JSON files in
data/ and regenerates the SVGs above. A per-codec breakdown lives in
wasmpeg's own COMPAT.md and
CORRECTNESS.md.