Cartesia

Usage guide

Setup, environment variables, and Python/JavaScript/Go usage examples.

CartesiaTTSModel

CartesiaTTSModel = typing.Literal['sonic-3.5', 'sonic-3', 'sonic-2', 'sonic-turbo']

CartesiaTTS

Cartesia Text-to-Speech provider. Synthesizes speech using Cartesia’s Sonic neural TTS models via streaming WebSocket, delivering low-latency audio in PCM format. Voices can be specified by catalog UUID or by a raw voice-embedding vector for voice-cloning use cases. Fine-grained generation controls (speed, emotion, volume) are available on sonic-3 and newer models.

Constructor

CartesiaTTS(*, api_key: 'str | None' = None, voice: 'Union[str, List[float], None]' = None, voice_id: 'Union[str, List[float], None]' = None, model: 'CartesiaTTSModel | str' = 'sonic-2', language: 'str' = 'en', sample_rate: 'int' = 24000, speed: 'Optional[float]' = None, volume: 'Optional[float]' = None, emotion: 'Optional[str]' = None, pronunciation_dict_id: 'Optional[str]' = None, max_buffer_delay_ms: 'Optional[int]' = None, enable_word_timestamps: 'Optional[bool]' = None, generation_config: 'Optional[GenerationConfig]' = None, **kwargs)

api_key

str | None

Cartesia API key for authentication. Falls back to the CARTESIA_API_KEY environment variable when omitted.

voice

Union[str, List[float], None]

Voice specification. Pass a str UUID from the Cartesia voice catalog (e.g. "f8f5f1b2-f02d-4d8e-a40d-fd850a487b3d"), or a list[float] embedding vector for voice cloning. Mutually exclusive with voice_id; voice takes precedence. Defaults to a built-in voice UUID when both are None.

voice_id

Union[str, List[float], None]

Alias for voice provided for backward compatibility. Accepts the same str | list[float] values. Ignored when voice is set.

model

CartesiaTTSModel | str

default:"sonic-2"

Cartesia TTS model identifier. Defaults to "sonic-2". Current models: "sonic-3.5", "sonic-3", "sonic-latest". Generation-config params (speed, emotion, volume) require sonic-3 or newer. Accepts a CartesiaTTSModel enum value or a plain string.

language

str

default:"en"

BCP-47 / ISO 639-1 language code for synthesis. Defaults to "en" (English). Cartesia Sonic supports 35+ languages including "fr", "de", "es", "pt", "zh", "ja", "hi", "it", "ko", "nl", "pl", "ru", "sv", "tr", and "ar".

sample_rate

int

default:"24000"

Requested PCM output sample rate in Hz. Cartesia supports 8000, 16000, 22050, 24000, 44100, and 48000 Hz. Defaults to 24000. Note: the output rate is fixed; a mismatch triggers a warning.

speed

Optional[float]

Playback speed multiplier in the range [0.6, 1.5]. 1.0 is normal speed. Overrides generation_config.speed when set. Defaults to None (provider default). Available on sonic-3 and newer.

volume

Optional[float]

Output volume multiplier in the range [0.5, 2.0]. 1.0 is normal volume. Overrides generation_config.volume when set. Defaults to None. Available on sonic-3 and newer.

emotion

Optional[str]

Emotion tag applied to the synthesized voice. Primary values: "neutral", "calm", "angry", "content", "sad", "scared". The full set contains 53 emotion variants. Overrides generation_config.emotion when set. Defaults to None. Available on sonic-3 and newer.

pronunciation_dict_id

Optional[str]

ID of a Cartesia pronunciation dictionary for custom phoneme rules. Overrides generation_config.pronunciation_dict_id when set. Defaults to None.

max_buffer_delay_ms

Optional[int]

Maximum buffering delay in milliseconds before audio is flushed, controlling the trade-off between streaming latency and audio quality. Overrides generation_config.max_buffer_delay_ms when set. Defaults to None.

enable_word_timestamps

Optional[bool]

When True, includes per-word start/end timing in the streamed response. Overrides generation_config.enable_word_timestamps when set. Defaults to False.

generation_config

Optional[GenerationConfig]

A GenerationConfig instance grouping all generation parameters as an alternative to flat kwargs. Flat parameters always take precedence over the corresponding GenerationConfig field. Defaults to None.

aclose

def aclose(self) -> 'None'

Release any resources held by the provider.

emit

def emit(self, event: 'T', *args: 'Any') -> 'None'

Emit an event, invoking all registered handlers with the given arguments. Handlers are called in registration order. Coroutine handlers are scheduled on the running event loop. If the emitter is closed, the call is a no-op.

event

required

The event to emit.

interrupt

def interrupt(self) -> 'None'

Interrupt and cancel the current synthesis for the active session, if any.

off

def off(self, event: 'T', callback: 'Callable[..., Any]') -> 'None'

Remove a previously registered handler for an event. If the handler is not registered for the event, the call is a no-op.

event

required

The event the handler was registered for.

callback

Callable[..., Any]

required

The handler to remove.

on

def on(self, event: 'T', callback: 'Callable[..., Any] | None' = None) -> 'Callable[..., Any]'

Register a handler for an event. Can be used directly by passing a callback, or as a decorator when callback is omitted.

event

required

The event to listen for.

callback

Callable[..., Any] | None

The handler to invoke when the event is emitted. If None, a decorator is returned that registers the decorated function.

returns

Callable[..., Any]

The registered callback when callback is provided, otherwise a decorator that registers and returns the function it wraps.

on_first_audio_byte

def on_first_audio_byte(self, callback: "'Callable[[int, int], Awaitable[None] | None]'") -> 'None'

callback

'Callable[[int, int], Awaitable[None] | None]'

required

A callable, optionally async, invoked with synthesis timing information when the first audio byte is emitted.

synthesize

def synthesize(self, text: 'Any', **kwargs: 'Any') -> 'None'

Synthesize the given text into speech. Implementations should convert the text to audio. The default behavior forwards the text to the active session for playback when one is present.

text

Any

required

The text to synthesize; coerced to a string if needed.

GenerationConfig

Per-request generation settings for CartesiaTTS. An alternative to supplying individual flat parameters directly to CartesiaTTS.__init__. When both a flat parameter and the corresponding GenerationConfig field are set, the flat parameter takes precedence.

Fields

speed

Optional[float]

Playback speed multiplier in the range [0.6, 1.5]. 1.0 is normal speed. Defaults to None (provider default).

emotion

Optional[str]

Emotion tag applied to the synthesized voice. Primary options include "neutral", "calm", "angry", "content", "sad", "scared". Defaults to None.

volume

Optional[float]

Output volume multiplier in the range [0.5, 2.0]. 1.0 is normal volume. Defaults to None.

pronunciation_dict_id

Optional[str]

ID of a Cartesia pronunciation dictionary to apply custom phoneme rules. Defaults to None.

max_buffer_delay_ms

Optional[int]

Maximum buffering delay in milliseconds before audio is flushed, used to control streaming latency vs. quality. Defaults to None.

enable_word_timestamps

bool

default:"False"

When True, the response will include word-level start/end timing. Defaults to False.

Python SDK

Core

Speech-to-text

LLM

Text-to-speech

Realtime models

Avatars

Turn detection

Voice activity detection

Noise cancellation

Usage guide

CartesiaTTSModel

CartesiaTTS

Constructor

aclose

emit

interrupt

off

on

on_first_audio_byte

synthesize

GenerationConfig

Fields

Usage guide

​CartesiaTTSModel

​CartesiaTTS

​Constructor

​aclose

​emit

​interrupt

​off

​on

​on_first_audio_byte

​synthesize

​GenerationConfig

​Fields

CartesiaTTSModel

CartesiaTTS

Constructor

aclose

emit

interrupt

off

on

on_first_audio_byte

synthesize

GenerationConfig

Fields