First-party researchAugust 1, 2026
13 audio encodes.3 browser engines.One repeatable test.
Measured file size, effective bitrate, encoding time, and real fixture loading across Chromium, Firefox, and WebKit for 13 WAV, FLAC, MP3, AAC, Ogg Vorbis, and Opus variants.
What the data says
Measured findings, without a fake quality score
FLAC reduced this PCM fixture by 43.6%
FLAC remained lossless at 6.5 MB. Compression varies with the source, so speech, silence, and dense music will produce different ratios.
128 kbps MP3 used 8.3% of the WAV bytes
The measured 960.8 KB file landed almost exactly on its constant-bitrate target. That says nothing by itself about perceptual quality.
64 kbps Opus was the smallest result
Opus 64 kbps produced 571 KB, or 5% of the WAV source. Compatibility remains part of the decision.
Measured output
Size and encoding results
File sizes are actual output bytes from the same 60-second source. Encoding time is the median of 3 local FFmpeg runs and should be used for relative context only.
| Variant | Actual file | Effective bitrate | Share of WAV | Median encode |
|---|---|---|---|---|
| WAV PCMpcm_s16le · .wav | 11.52 MB | 1536.0 kbps | 100.0% | 119 ms |
| FLACflac · .flac | 6.5 MB | 866.1 kbps | 56.4% | 362 ms |
| MP3 128 kbpsmp3 · .mp3 | 960.8 KB | 128.1 kbps | 8.3% | 971 ms |
| MP3 192 kbpsmp3 · .mp3 | 1.44 MB | 192.2 kbps | 12.5% | 1.02 s |
| MP3 320 kbpsmp3 · .mp3 | 2.4 MB | 320.3 kbps | 20.9% | 1.06 s |
| AAC 128 kbpsaac · .m4a | 974 KB | 129.9 kbps | 8.5% | 1.74 s |
| AAC 192 kbpsaac · .m4a | 1.45 MB | 193.9 kbps | 12.6% | 1.94 s |
| AAC 256 kbpsaac · .m4a | 1.94 MB | 258.1 kbps | 16.8% | 2.04 s |
| Ogg Vorbis 128 kbpsvorbis · .ogg | 710.1 KB | 94.7 kbps | 6.2% | 1.04 s |
| Ogg Vorbis 192 kbpsvorbis · .ogg | 1.07 MB | 142.4 kbps | 9.3% | 1.03 s |
| Opus 64 kbpsopus · .opus | 571 KB | 76.1 kbps | 5.0% | 594 ms |
| Opus 96 kbpsopus · .opus | 735.3 KB | 98.0 kbps | 6.4% | 656 ms |
| Opus 128 kbpsopus · .opus | 930.2 KB | 124.0 kbps | 8.1% | 764 ms |
No results match this filter.
Real file loading
Browser-engine compatibility matrix
A pass means the generated file reached loadedmetadata with the expected duration. These are Playwright engine builds on one Windows host, not branded desktop or mobile browser certifications.
Chromium149.0.7827.55
Firefox151.0
WebKit26.5
| Generated file | Chromium | Firefox | WebKit |
|---|---|---|---|
| WAV PCM | Loadedmaybe | Loadedmaybe | Loadedprobably |
| FLAC | Loadedprobably | Loadedmaybe | Timed outprobably |
| MP3 128 kbps | Loadedprobably | Loadedmaybe | Loadedprobably |
| MP3 192 kbps | Loadedprobably | Loadedmaybe | Loadedprobably |
| MP3 320 kbps | Loadedprobably | Loadedmaybe | Loadedprobably |
| AAC 128 kbps | Loadedprobably | Loadedprobably | Loadedprobably |
| AAC 192 kbps | Loadedprobably | Loadedprobably | Loadedprobably |
| AAC 256 kbps | Loadedprobably | Loadedprobably | Loadedprobably |
| Ogg Vorbis 128 kbps | Loadedprobably | Loadedprobably | Did not loadno MIME claim |
| Ogg Vorbis 192 kbps | Loadedprobably | Loadedprobably | Did not loadno MIME claim |
| Opus 64 kbps | Loadedprobably | Loadedprobably | Did not loadno MIME claim |
| Opus 96 kbps | Loadedprobably | Loadedprobably | Did not loadno MIME claim |
| Opus 128 kbps | Loadedprobably | Loadedprobably | Did not loadno MIME claim |
Reproduce and critique it
Methodology and limits
- Generate one deterministic 60-second, 48 kHz, stereo, 16-bit PCM source fixture.
- Encode every variant three times with the listed FFmpeg codec and target bitrate.
- Measure final bytes, effective bitrate, and the median wall-clock encode time on one test host.
- Serve each produced file to current Playwright Chromium, Firefox, and WebKit engines and require loaded metadata from the real file.
- Treat browser results as a dated engine test, not a promise for every operating system, device, or vendor build.
Deterministic synthetic multi-tone and low-level noise fixture generated by FFmpeg. It is designed for repeatability, not to represent every voice or music recording.
- Duration
- 60 seconds
- Source
- 48 kHz, 16-bit, stereo
Encoding speed is hardware-dependent. File bytes and browser load outcomes are the more portable observations.
- CPU
- AMD Ryzen 7 4800H with Radeon Graphics
- FFmpeg
- 2025-07-17-git-bc8d06d541-full_build-www.gyan.dev
Data is available under CC BY 4.0. Cite AudioFetcher, this URL, and the August 1, 2026 benchmark date.
Use the numbers
Turn the benchmark into a practical decision
Estimate your own file sizeChange duration, bitrate, sample rate, channels, and file count.
Compare MP3 and WAVChoose a working format or a smaller delivery copy.
Understand codecs and containersLearn what common extensions actually contain.
Open 30+ local audio toolsProcess files in your browser without sending media to our server.
Questions
How to read this benchmark
Which audio format made the smallest file?
Opus 64 kbps produced the smallest file in this fixture at 571 KB. That result does not mean it is the best choice for every source or playback target.
Does a smaller file prove better audio quality?
No. This benchmark measures size, bitrate, encoding time, and whether a browser engine loaded the fixture. It does not assign a subjective quality score.
Do the browser results guarantee support on my device?
No. They record one dated test of Playwright browser-engine builds on one Windows host. Operating-system codecs, vendor builds, and older devices can behave differently.
Can I reuse the benchmark data?
Yes. Download the CSV or JSON and cite this page with the benchmark date and methodology.