Where WAV files come from
WAV holds uncompressed audio, so it is the default for equipment where quality matters more than file size. Portable field recorders from brands like Zoom, Tascam and Sony record WAV by default. Recordings made through an audio interface in Audacity, Adobe Audition, Reaper or another audio editor are usually saved or exported as WAV. Researchers recording interviews, journalists using dedicated recorders, and podcast producers working from raw tracks all end up with WAV files.
Call centers and phone systems are another common source. Many of them export recorded calls as WAV, often at a low sample rate.
Large WAV files, and why size does not change accuracy
Uncompressed audio is big. A CD-quality stereo WAV (44.1 kHz, 16-bit) uses a little over 10 MB per minute, and a 48 kHz, 24-bit file is larger still. An hour-long interview can easily pass half a gigabyte.
Before recognition, every file is converted to a single mono track at 16 kHz, which is the rate the speech model works at. Recording at 48 kHz or 96 kHz with 24-bit depth is useful for editing and mixing, but it does not make the transcript more accurate than a clean 16-bit recording of the same voice.
Uploads are sent in parts and Retry continues where it stopped, so large files survive an unstable connection. If a WAV is above your plan's file size limit, convert it to FLAC (lossless, roughly half the size) or to a high-bitrate MP3. Both are accepted and the transcript will be the same.
Stereo, dual-mono and multitrack recordings
Interview setups often put one lavalier microphone on the left channel and another on the right. Some recorders add a backup safety track at a lower level. When you upload a stereo or multichannel WAV, all channels are mixed into one before transcription, so both speakers are included.
If your recorder saved each microphone as a separate WAV file, each file is transcribed on its own. To get one transcript for the whole conversation, mix the tracks into a single WAV in your editor first. On paid plans, speaker labels then separate the voices in that combined file.
Getting a clean result from field and phone recordings
Recording levels matter more than format. A voice that clips, where the peaks hit the top of the meter and distort, loses detail that no software can recover. A recording that is quiet but clean usually transcribes fine. Recorders that write 32-bit float WAV avoid clipping, and those files are accepted like any other WAV.
Telephone audio is usually recorded at 8 kHz, which cuts off the higher frequencies that help distinguish sounds like s and f. These recordings still transcribe, but expect more errors than with a studio or field recording, especially with accents, fast speech or a poor line. Upsampling an 8 kHz file to 44.1 kHz does not restore the missing detail, so upload it as it is.
The language is detected automatically, or you can set it per file. Speech in 90+ languages is supported, including conversations that mix Chinese and English.
WAV to text exports for research and production
Pick the export that matches the next step of your work. Timestamps and speaker names can be switched on or off in each one.
- DOCX: interview transcripts for qualitative analysis, annotation or review.
- CSV: one row per line with start times, ready for a spreadsheet or analysis software.
- SRT or VTT: subtitle timing when the WAV is the production audio for a video edit.
- TXT or PDF: plain copies for archives, search or sharing.