Inworld TTS
InworldTTS targets the Inworld TTS HTTP endpoint (POST https://api.inworld.ai/tts/v1/voice:stream). The response is NDJSON — one JSON object per line of the form {"result": {"audioContent": "<base64>", "timestampInfo": ...}} — and the provider yields the base64-decoded audio chunks as they arrive.
The default model is inworld-tts-2 (sub-200 ms time-to-first-audio, 100+ languages with mid-utterance switching, natural-language voice steering). Pass model="inworld-tts-1.5-max" to fall back to the prior generation when you need temperature control.
The default audio output is PCM_S16LE @ 16 kHz so chunks drop straight into the Patter pipeline without transcoding.
Install
getpatter[inworld] extra adds aiohttp>=3.10 for streaming the NDJSON body. (TypeScript uses native fetch and needs no extra dependency.)
Authentication
The Inworld dashboard issues a Base64 token that is already in the form
expected by the
Authorization: Basic <token> header — paste it into
INWORLD_API_KEY as-is. Do not re-encode it.If you only have the raw API key string, base64-encode "<api_key>:"
(note the trailing colon) yourself before passing it in.Usage
Customising delivery mode (TTS-2)
deliveryMode controls how expressive the TTS-2 voice is. Use EXPRESSIVE for warm conversational agents, STABLE when you want consistent, predictable prosody (e.g. for IVR-style read-backs), and BALANCED for the middle ground.
Switching to TTS-1.5 for temperature control
When you need sampling-temperature control (e.g. for more variation across multi-turn conversations), drop down to inworld-tts-1.5-max:
Models
Options
Low-level usage
If you want the streaming generator without going through the pipeline-mode wrapper:Pricing
The default rate inpricing.py is $0.020 / 1k characters for inworld-tts-2, with $0.025 / 1k for the TTS-1.5 family. These are placeholder defaults — verify against your current Inworld platform tier and override per-project via Patter(pricing={...}) if needed.
