WAV to text for serious recordings
WAV is the format of people who care about their audio: studio sessions, field recordings, voice-over takes, courtroom and archive digitizations. Those files are big — a single hour of uncompressed audio can pass 600 MB — which makes “upload it to our server” a terrible workflow.
This tool skips the upload entirely. The WAV is decoded in your browser and transcribed on your own hardware, so a huge file starts processing immediately instead of crawling up your connection. Your files are never uploaded — everything runs on your device.
Checking what this browser supports…
How to transcribe a WAV file
-
Drop in the WAV
Any sample rate or channel count works — the tool downmixes and resamples automatically before transcription.
-
Processing starts instantly
No upload wait, even for very large files. The transcript appears progressively, with progress shown against the recording’s length.
-
Refine and export
Correct technical vocabulary inline, then export TXT for documents, SRT/VTT for subtitling, or JSON with precise segment timings.
Built for heavyweight files
No upload bottleneck
A 600 MB session file would take ages to upload. Reading it locally takes seconds — the only real work is the transcription itself.
Any sample rate in
44.1 kHz CDs, 48 kHz video audio, 96 kHz masters, 8 kHz telephone archives: everything is resampled correctly under the hood.
Masters stay yours
Unreleased recordings and client sessions never touch a third-party server. What you recorded stays exactly where you put it.
WAV transcription questions
- Is there a file-size limit?
- No hard limit is enforced. Very large files are bounded by your device’s memory; the audio is processed in windows to keep usage steady during transcription itself.
- Do high sample rates improve the transcript?
- Not really — speech models listen at 16 kHz, and your file is resampled to that automatically. High sample rates don’t hurt, but they don’t add accuracy either.
- Stereo, mono, multi-channel?
- All fine. Channels are averaged down to mono before transcription, which is what speech recognition expects.
- What model does the recognition?
- OpenAI’s open-source Whisper model, executed locally in your browser by our open-source engine. Nothing about your audio is sent out to run it.
- Can I transcribe several takes in a row?
- Yes — drop the next file as soon as the previous transcript is done (or cancel mid-way). The model stays loaded, so follow-up files skip the download and start faster.