Changelog

0.0.1b2 — 2026-10-09

• Removed public seed controls from buffered, streaming, and WebSocket speech requests. The server manages the seed internally.

0.0.1b1 — 2026-10-09

• Reduced speech voice selection to a voice ID or zero-shot reference-audio cloning.
• Removed voice descriptions and reference transcripts from native and patched speech requests.
• OpenAI instructions now warns without changing audio.

0.0.1b0 — 2026-08-29

• Added request-time voice_description support with or without reference audio.
• Added canonical reference_transcript support and retained reference_text as a compatibility alias.
• Added explicit validation for incompatible description-and-transcript requests.
• Added the same voice-modelling fields to the OpenAI and ElevenLabs patches.

0.0.1a0 — 2026-08-23

• Added synchronous and asynchronous clients with TW aliases.
• Added speech creation, HTTP and SSE streaming, and WebSocket input streaming.
• Added public voices, models, capabilities, and project usage.
• Added inline voice conditioning and the cfg_value control.
• Added OpenAI and ElevenLabs TTS patches with provider pass-through.
• Added typed errors, warnings, patch handles, and provider-version checks.