TTS Output Audio Format - Cartesia Docs

This page explains how to configure output_format for TTS responses (container, encoding, and sample_rate). In general, use a consistent encoding and sample rate across your audio pipeline (telephony, playback, and storage) to avoid unnecessary transcoding and quality loss. If you’re saving audio samples, we recommend using the Text-to-Speech (Bytes) API with output_format.container: "wav" or output_format.container: "mp3" so audio players can automatically detect the encoding and sample rate.

Reference

output_format.container

string

The container format for the audio output.Available options: raw, wav, mp3. Only the Bytes endpoint supports all container formats; our other endpoints (SSE, Websockets) only support raw.

output_format.encoding

string

The encoding of the output audio. Available options: pcm_f32le, pcm_s16le, pcm_mulaw, pcm_alaw.

output_format.sample_rate

number

The sample rate of the output audio. Remember that to represent a given signal, the sample rate must be at least twice the highest frequency component of the signal (Nyquist theorem).Available options: 8000, 16000, 22050, 24000, 44100, 48000.

`output_format` for RAW (PCM) Audio

When using raw audio, it is important to match the encoding and sample rate with your output device with the output_format parameter.

Encoding	Bit depth	Commonly used for	Pair with sample rate
`pcm_mulaw`	8-bit compressed	North American / Japanese telephony (G.711μ), Twilio	8000
`pcm_alaw`	8-bit compressed	European / international telephony (G.711A)	8000
`pcm_s16le`	16-bit int	Most voice agent platforms, e.g. LiveKit and Pipecat	16000
`pcm_f32le`	32-bit float	Most browsers	44100 or 48000

Telephony

North America and Japan

Many customers send their audio output over Twilio. All audio sent over Twilio is transcoded to µ-law encoding with an 8 kHz sample rate.

{
  "container": "raw",
  "encoding": "pcm_mulaw",
  "sample_rate": 8000
}

Europe, India, and others

The standard for European and international telephone networks (G.711A) is 8-bit A-law compressed PCM with an 8 kHz sample rate.

{
  "container": "raw",
  "encoding": "pcm_alaw",
  "sample_rate": 8000
}

Voice agent platforms

Many voice agent platforms use pcm_s16le at a 16 kHz sample rate in their pipeline. If speech plays alright, then that probably means you’re using the right output format.

Web browsers

When playing audio through the Web Audio API, use pcm_f32le to match AudioContext. The sample rate is not fixed: it defaults to whatever the user’s output hardware reports, commonly 48 kHz but sometimes 44.1 kHz. Read it from AudioContext.sampleRate:

const audioContext = new AudioContext();
console.log(audioContext.sampleRate); // e.g. 48000

Match your output_format.sample_rate to this value. If it differs, the browser resamples on playback, which adds latency and can degrade quality. You can request a specific rate with new AudioContext({ sampleRate: 24000 }), but browsers may error depending on the platform.

Audio CD quality

Standard audio CDs are encoded as pcm_s16le at a 44.1 kHz sample rate.

{
  "container": "raw",
  "encoding": "pcm_s16le",
  "sample_rate": 44100
}

This performs well for consumer digital audio setups.

Documentation Index

​Reference

​output_format for RAW (PCM) Audio

​Telephony

​North America and Japan

​Europe, India, and others

​Voice agent platforms

​Web browsers

​Audio CD quality