Audio formats, sample rates and file sizes from TTS tools

Many TTS tools can give you more than one kind of audio file. The right one depends on where it will be played. Start with the place the audio is going, not with the tool.

The terms, in plain words

Format is the kind of file: MP3, WAV, Ogg and so on. Sample rate counts the snapshots of the sound taken each second, measured in hertz (Hz), when the sound becomes digital data. A WAV file stores this in a header section, along with the bit depth and the number of channels.¹ Channels means mono (one) or stereo (two).

Compression shrinks the file. An MP3 is made with lossy compression, which throws away part of the data in return for a much smaller file.² Lossy is the word to watch. Detail that's thrown away stays gone, so keep a master copy in a larger format if you may edit the audio later.

WAV is the main format for uncompressed audio on Windows.¹ Be aware that a .wav file isn't always uncompressed: all WAV files, even those using MP3 compression, use the .wav extension.¹ If a phone system rejects a WAV, the encoding inside may be the problem, not the extension.

What one maker offers

Amazon Polly's documentation, as read on 2 November 2026, lists these output types for audio: mp3, ogg_vorbis, ogg_opus, mu-law, a-law and pcm.³

If you're loading greetings onto a business phone line, look for the format your phone system's manual asks for, then see whether your TTS tool offers it. Follow the manual for your own model rather than guessing.

Matching the file to the job

• Phone system or hold messages. Open the system's manual or support page and find the accepted file type and sample rate. If it names a format your TTS tool offers, pick that one. A mismatch may mean the upload is rejected.
• Video editor. Editors handle common formats, but check the editor's own help page. Keeping a WAV master avoids repeated lossy compression when you edit.
• Website or e-learning course. Check which file types the platform accepts and any size limit it sets. An MP3 file is compressed, so it's smaller than an uncompressed one.²
• Mono or stereo. Google's documentation for inserting audio into SSML says single channel is preferred, though stereo is acceptable.⁴ Other tools set their own rules.

Before you hand the file over

• Play the file in the exact place it will be used, not just on your computer.
• Keep the original export and note the settings you used.
• Check the tool's documentation for changes, since formats and defaults move. Google's page, for example, marks its WAV option for audio insertion as deprecated.⁴

The destination decides the format. A short test file will answer your question faster than any guide. To see how the audio gets made in the first place, read how a computer reads text aloud.