wav → text · free · no signup
WAV to text, without the giant upload.
WAV files are huge. Your browser shrinks them to exactly what the speech model reads before a single byte is uploaded.
// short answer
To convert WAV to text, drop the .wav file into the box above. Your browser downsamples it to 16 kHz mono, about 1.8 MB per minute, then uploads and transcribes it. On the free tier that fits roughly 8 minutes of WAV audio. Unlimited ($9/month) takes files up to 2 hours and 50 MB, 30 hours of audio a month, no daily limit, with speaker labels.
Updated 2026-09-17
Why the file gets smaller and the transcript doesn't get worse
Whisper converts every input to 16 kHz mono before it looks at a single sample. Anything richer (44.1 or 48 kHz, stereo, 24-bit) is data the model throws away.
So the page does that conversion in your browser first. In our own test, a 1 minute 47 second, 48 kHz stereo WAV went from 19.6 MB to 3.4 MB, carrying everything the model actually reads.
How long a WAV fits for free
After downsampling, WAV is still uncompressed: about 1.8 MB per minute. The free upload cap is 15 MB, so about 8 minutes fit. Compressed formats go much further. The same 30-minute recording as a 48 kbps MP3 is about 10 MB.
For a WAV longer than 8 minutes, convert it first:
ffmpeg -i recording.wav -vn -ac 1 -b:a 48k output.mp3Where WAV files come from
- Field and handheld recorders. Most default to WAV. Record interviews at 16 or 44.1 kHz mono. Higher settings only make the files bigger.
- Audacity, Audition, Logic and other editors. Export as MP3 instead of WAV if all you need is a transcript.
- Call recorders and dictation software. Often WAV at low sample rates, which are already small.
Free vs Unlimited
| Free | Unlimited | |
|---|---|---|
| Price | $0 | $9/month, cancel anytime |
| Files per day | 3 | No daily limit |
| Longest file | 15 min on screen, 30 min by email | 2 hours |
| Upload size | 15 MB | 50 MB |
| Audio per month | — | 30 hours |
| Speaker labels | No | Yes |
| Account | None | None, Stripe checkout |
FAQ
Does downsampling to 16 kHz hurt accuracy?+
No. The model resamples everything to 16 kHz mono itself. Doing it in the browser first changes the upload size, not the transcript.
My interview has each person on a separate stereo channel. Is that used?+
No. Channels are averaged to mono before upload. On Unlimited, speakers are separated by voice instead, and labelled Speaker 1, Speaker 2 and so on.
Why was my 20-minute WAV refused on the free tier?+
Twenty minutes of 16 kHz mono WAV is about 36 MB, over the 15 MB free cap. Convert it to MP3 with the command on this page, or use Unlimited (50 MB cap).
Does it matter if my WAV is 24-bit or 32-bit float?+
No. The browser decodes any standard PCM WAV and converts it to 16-bit before upload.