Media conversion compatibility profile — six real FFmpeg cases

Measured 2026-08-16 · ffmpeg version 8.1.2 Copyright (c) 2000-2026 the FFmpeg developers; native media-engine profile 1.0.0. Dataset FH-MEDIA-SYNTH-2026-08: two-second synthetic motion/color video and generated audio tones; no visitor files or personal data.

Question: Do all six media tools produce the exact container, codecs and duration behaviour published on their pages?

Findings

All six fixtures passed signature inspection, output ffprobe, a complete output decode, promised-codec verification and the duration-tolerance check.
The largest observed duration difference was 0.020 seconds — inside the published 0.250 second floor.
Compatibility conversion is not synonymous with compression: M4A to MP3 grew from 31,930 to 50,199 bytes because the fixed MP3 delivery bitrate exceeds the tiny synthetic AAC source.
Every fixture and expected output is downloadable from its tool page, and the lab exports this matrix as CSV and JSON.

The data

Case	Tool	Input codecs	Output codecs	Input SHA-256	Output SHA-256	Input bytes	Output bytes	Size change	Duration delta	Elapsed ms	Validation
FH-MEDIA-01	webm-to-mp4	vp9 + opus	h264 + aac	834490a1a8b57748d371d0da60aa553b69f8ccfd3753093a1f8eff657180dae2	69b3ec972990330e85df6b4d2e8875de339877421fa0753e3cac7f6214c5ea46	93,031	89,611	-4%	0.008 s	162	pass
FH-MEDIA-02	mov-to-mp4	h264 + pcm_s16le	h264 + aac	6cd69d1d4d42f9652fc4737f0b491f68c96c7fcebbd2d26e4c5d0ece228d35cc	667430357cbeaab2e7d0b7272003aadb38756dcf88627541273d5d258814ee06	273,388	84,697	-69%	0 s	132	pass
FH-MEDIA-03	mp4-to-mp3	h264 + aac	mp3	076cee70b78122b5d9197f57a1eb5fa1cd2873e167ce437bb0af831b6697b973	0b400f5efcaefcb79da7bfe5d284a345b5aa238b84d81e64d297645bd86acf27	89,106	49,581	-44%	0.005 s	87	pass
FH-MEDIA-04	webm-to-mp3	vp9 + opus	mp3	834490a1a8b57748d371d0da60aa553b69f8ccfd3753093a1f8eff657180dae2	e660d5eb3a9946084a018762e150d5237abd97cc604bc3e771ef0023f4cefb43	93,031	49,581	-47%	0.008 s	85	pass
FH-MEDIA-05	m4a-to-mp3	aac	mp3	ad072f89c019d2b7aba7c4772c121f737c9f11c30c62dc2bfca3a78c3b66534a	2d57fcc5831817606985c7e3e364cebcc53fc3928662c705ba89fa6b96202746	31,930	50,199	+57%	0.02 s	84	pass
FH-MEDIA-06	wav-to-mp3	pcm_s16le	mp3	b192ea5a2efce8baebc5837afd2fb19c1c2436b0b933814b602db9ccc0cb8ec5	4b61b96caf5f60a1f8bb9a6b06e6f97528c046e5d70f3d686489a111206ad448	192,078	49,581	-74%	0 s	97	pass

⬇ Download raw data (CSV) ⬇ Download report (JSON)

Method

Generate five copyright-free fixtures with FFmpeg: VP9/Opus WebM, H.264/PCM MOV, H.264/AAC MP4, AAC M4A and PCM WAV.
Run the six public task profiles through the same media-engine.js module used by the server — no hand-written substitute command.
Inspect every input and finished output with ffprobe, then decode every output stream from start to finish with FFmpeg in error-fatal mode. Require the declared stream, MP4=H.264/yuv420p (+ AAC when audio exists), MP3=audio-only MP3, and a duration difference no greater than 250 ms or 1%.
Record exact bytes, SHA-256 checksums, codecs, duration deltas and elapsed time in data/media-benchmark.json. The generator overwrites the dataset rather than asking an editor to copy numbers.

Limitations

Small synthetic fixtures prove the fixed profiles and validation path, not subjective quality on every camera, codec, HDR mode or damaged timestamp layout.
Encoding speed is hardware-dependent; elapsed milliseconds are published for reproducibility but are not a promise about visitor wait time.
One primary stream per task. Alternate audio, subtitles, chapters, attachments and metadata are deliberately outside this compatibility profile.

Takeaway

These results support a narrow, checkable claim: profile 1.0.0 produces the documented codecs and timing on the six published synthetic cases. They do not support a universal quality score, so the tools report structural pass/fail evidence and disclose lossy changes instead.

← All Labs reports · standards in the testing methodology