Providers
inworld-tts-1.5-max
The Inworld text-to-speech model inworld-tts-1.5-max, with every setting it accepts, the allowed value and the default.
The highest-quality 1.5 model, and the one to pick when delivery matters more than latency. Uses temperature for variation.
Reached through the LiveKit Inference gateway.
The two generations steer variation differently: the 1.5 models use temperature, and TTS-2 uses deliveryMode and ignores temperature. Setting the wrong one for your model is not an error; it is ignored.
Settings#
Every setting is optional. Leave one out and OnCore sends nothing for it, so the Default applies. If invalid is what the provider does with a value it does not accept.
| Setting | Type | Allowed | Default | If invalid |
|---|---|---|---|---|
speakingRate | number | above 0.5, up to 1.5 | 1.0 | Refused, call fails |
temperature | number | 0 to 2; 0.6 to 1.1 recommended | 1.1 | Refused, call fails |
timestampType | string | WORD or CHARACTER | no timestamps | Refused, call fails |
applyTextNormalization | string | ON, OFF | ON | Refused, call fails |
Overriding on a single call#
Return this from your assistant-request webhook to change these settings for one call only. The stored voice is left untouched, and the next call starts from it again.
{
"data": {
"overrides": {
"voice": {
"settings": {
"speakingRate": 1,
"temperature": 0.8,
"timestampType": "WORD",
"applyTextNormalization": "ON"
}
}
}
}
}Send only the keys you want to change: anything you leave out keeps the value the assistant already has. Add "id" beside "settings" to swap the whole voice for another one from your catalogue at the same time.
timestampType: Set to WORD or CHARACTER this turns on word-level or character-level timing for the synthesized audio. Left unset, no timing is produced. TIMESTAMP_TYPE_UNSPECIFIED is accepted and behaves the same as leaving it unset.
Next steps#
- inworld-tts-1.5-mini: the next Inworld model.
- inworld-tts-2: the previous Inworld model.
- Providers introduction: per-call overrides, and how a wrong value behaves across providers.