Usage guide
Setup, environment variables, and Python/JavaScript/Go usage examples.
RimeTTS
Text-to-speech with Rime’s conversational voices, including the Mist and Arcana models. Good for natural, expressive speech with fine pronunciation control. Authenticates with RIME_API_KEY from your environment, or an explicitapi_key.
Fields
str
default:"UNSET"
Rime voice to speak with, e.g. “river”. The voice must be available for the chosen model. Defaults to ‘river’.
str
default:"UNSET"
Rime model to synthesize with, e.g. “mist”, “mistv2” or “arcana”. Defaults to ‘mist’.
str
default:"UNSET"
Three-letter language code, e.g. “eng” or “spa”. Defaults to ‘eng’.
int
default:"UNSET"
Output audio sample rate in Hz, e.g. 8000, 16000, 22050 or 24000. Defaults to 24000.
float
default:"UNSET"
Speaking rate multiplier; values below 1.0 speak faster and above 1.0 speak slower. Defaults to 1.0.
bool
default:"UNSET"
Skip some text normalization to return audio sooner. Defaults to True.
bool
default:"UNSET"
Insert pauses written in angle brackets in the text, e.g. “<500>” for 500 ms. Defaults to False.
bool
default:"UNSET"
Read text in curly brackets as phonemes, for custom pronunciations. Defaults to False.
str | None
default:"UNSET"
Comma-separated speed values applied to bracketed sections of the text, e.g. “0.5, 3”. Defaults to None (not set).
Optional[float]
default:"UNSET"
Sampling temperature controlling how varied the delivery is. Arcana model only. Defaults to None (not set).
Optional[float]
default:"UNSET"
Nucleus sampling cutoff, 0 to 1. Arcana model only. Defaults to None (not set).
Optional[float]
default:"UNSET"
Penalty that discourages repeated sounds or words. Arcana model only. Defaults to None (not set).
Optional[int]
default:"UNSET"
Maximum number of output tokens per request, limiting audio length. Arcana model only. Defaults to None (not set).