Why WAV
WAV is uncompressed audio, so what you upload is exactly what the voice model hears. Converting an MP3 to WAV doesn't bring back detail the MP3 threw away, but a clean MP3 converted to WAV clones perfectly well. What matters more is that the recording itself is clean: one speaker, no music, no echo.
With Audacity (Windows, Mac, Linux)
Audacity is free and does the conversion and the trimming in one go.
- Open the file with File › Open. MP3 and WAV open straight away; M4A, AAC and some others need Audacity's free FFmpeg library, which Audacity will offer to help you install.
- Choose the best 5 to 15 seconds. Drag across a stretch of clear, natural speech, then use Edit › Remove Special › Trim Audio to keep only that part.
- Make it mono if it's stereo (optional): Tracks › Mix › Mix Stereo Down to Mono.
- Export with File › Export Audio, choose WAV as the format, and keep the default 16-bit PCM encoding.
With ffmpeg (any computer, from the command line)
If you have ffmpeg installed, one command converts a file:
ffmpeg -i sample.m4a sample.wav
To keep only part of it, say 12 seconds starting 2 seconds in, and make it mono at the same time:
ffmpeg -i sample.mp3 -ss 2 -t 12 -ac 1 sample.wav
On a Mac, without installing anything
macOS comes with a converter called afconvert. In Terminal:
afconvert -f WAVE -d LEI16 sample.m4a sample.wav
The Music app can do it too: in Music › Settings › Files › Import Settings, choose WAV Encoder, then select the recording in your library and use File › Convert › Create WAV Version.
What about online converters?
They work, but they mean uploading a recording of someone's voice to a website you know nothing about. Voices are personal, so a tool on your own computer is the better choice, especially for anyone else's voice.
Or don't convert at all
The Record in browser button in abcVoiceover records straight from your microphone and saves WAV automatically, on a computer or a phone. If you haven't recorded the sample yet, that's the easiest route. See how to record a voice sample.
If the upload is refused
- "Only .wav files can be uploaded for now": the file is still an MP3 or M4A, perhaps renamed. Convert it with one of the tools above; renaming the file isn't enough.
- "Voice clips need at least 3s of clear speech": the clip is too short. Aim for 5 to 15 seconds.
- "Audio is too long": the maximum is 10 minutes, but you only need a few seconds. Trim it.