How fast is browser audio conversion?
On our test computer, a roughly 25 MiB WAV converted to 192 kbps MP3 in about 1.4 seconds. A roughly 100 MiB WAV took about 3.8 seconds. These are local development measurements from one computer, not a speed guarantee.
The actual runs
| Input | Engine cache | Wall time | Output |
|---|---|---|---|
| 25 MiB WAV | cold | 1.41 s | MP3, 192 kbps |
| 25 MiB WAV | warm | 1.35 s | MP3, 192 kbps |
| 100 MiB WAV | cold | 3.86 s | MP3, 192 kbps |
| 100 MiB WAV | warm | 3.84 s | MP3, 192 kbps |
What we measured
Five things make this reproducible: a generated 440 Hz sine wave, stereo 16-bit PCM at 44,100 Hz, our pinned FFmpeg 5.1.10 audio engine, fixed 192 kbps MP3 export, and a recorded input checksum. A fresh browser context starts each size’s cold run; a second run uses that context’s HTTP cache. Workers are recreated on every conversion.
Host: Apple M3; darwin 25.0.0. Browser: Chromium 153.0.8010.12. Measured 2026-09-17. Each condition has one run; four runs total. Output codec and duration were checked with native ffprobe.
What a slow connection changes
We also tested a five-second MP4 with Chromium network emulation at 1.6 Mbps and 150 ms latency. Each of three fresh browser contexts made a cold conversion and then a repeat conversion using its HTTP cache. Desktop CPU speed was unchanged. Timing starts at the Convert click, after the page loads.
| Trial | Engine cache | Wall time |
|---|---|---|
| 1 | cold | 15.04 s |
| 1 | warm | 0.83 s |
| 2 | cold | 14.95 s |
| 2 | warm | 0.82 s |
| 3 | cold | 14.92 s |
| 3 | warm | 0.82 s |
The cold conversions took about 15 seconds; repeat conversions about 0.8 seconds. The engine is downloaded once and cached when the browser allows it. These are simulated network conditions, not measurements from a phone or a public CDN. Download the six network-test results and methodology.
What these numbers do not show
The website was served over local loopback. A public cold download of the 2.65 MB processing engine can take much longer. We did not measure peak memory, perceptual music quality, real phone performance, or population failure rates. A clean synthetic tone is easy to encode and does not represent all recordings.
250 MB, 500 MB and 1 GB inputs exceed this release’s limit and were not processed. Mobile browser emulation is a layout test, not evidence about a physical phone’s memory or battery.
Download raw measurements and input hashes
Try your own recording
Convert WAV to MP3. Your file stays on your device. If a particular file fails, a smaller clip may succeed.