v0.2.12

`speech-to-speech` 0.2.12 is the final planned release in the 0.2.x line before
the next round of larger changes. It brings smarter turn-taking, WebRTC support,
direct audio input for audio-capable LLMs, more complete OpenAI Realtime protocol
behavior, and a significantly improved browser demo.

### Highlights

- **Smarter endpointing with Smart Turn v3.2.** Realtime mode now enables the
  quantized CPU Smart Turn model by default to distinguish completed turns from
  mid-thought pauses while speculative STT and LLM work continues. Use
  `--no_smart_turn` to retain Silero-only endpointing. ([#192](https://github.com/huggingface/speech-to-speech/pull/192))
- **WebRTC transport for the OpenAI Realtime API.** Install the new `webrtc`
  extra to use SDP negotiation at `POST /v1/realtime/calls`, RTP audio, and the
  `oai-events` data channel alongside the existing WebSocket transport. Both
  transports share the same pipeline pool, event dispatch, cancellation, and
  interruption behavior. ([#352](https://github.com/huggingface/speech-to-speech/pull/352))
- **Direct audio input for audio-capable LLMs.** Run with `--stt none`, the
  Chat Completions backend, and an explicitly selected audio-capable model to
  send completed VAD audio directly to the model. Audio history, cancellation,
  tool calls, usage accounting, and provider failures are handled transactionally.
  ([#298](https://github.com/huggingface/speech-to-speech/pull/298))
- **Optional remote-LLM proxy.** `--enable_llm_proxy` exposes the configured
  remote backend through the matching `/v1/responses` or
  `/v1/chat/completions` path, with streaming passthrough and usage accounting.
  The engine deliberately does not authenticate or throttle these routes, so
  standalone deployments must keep them on a trusted network or behind an
  authenticated gateway. ([#368](https://github.com/huggingface/speech-to-speech/pull/368))

### Realtime, demo, and backend improvements

- The browser demo can select microphone and speaker devices, start with a
  configurable assistant greeting, forward signed-in Hugging Face identity for
  session allocation, show user-turn lifecycle feedback, and replay the exact
  captured user audio locally. ([#376](https://github.com/huggingface/speech-to-speech/pull/376),
  [#389](https://github.com/huggingface/speech-to-speech/pull/389),
  [#390](https://github.com/huggingface/speech-to-speech/pull/390))
- Qwen3-TTS GGML users can select quantization, load local talker and codec GGUF
  files, configure voice-reference caching, and reuse precomputed `.spk` and
  `.rvq` references. ([#403](https://github.com/huggingface/speech-to-speech/pull/403))
- Successful `session.update` requests now receive the expected
  `session.updated` event with the effective session configuration.
  ([#413](https://github.com/huggingface/speech-to-speech/pull/413),
  [#417](https://github.com/huggingface/speech-to-speech/pull/417),
  [#418](https://github.com/huggingface/speech-to-speech/pull/418))
- Language prompting now covers every language reported by the bundled STT
  handlers, including the full default Parakeet TDT language set.
  ([#395](https://github.com/huggingface/speech-to-speech/pull/395))
- A new guide documents a local Gemma 4 12B Realtime setup on Apple Silicon,
  including native-audio bypass, memory tuning, cancellation, barge-in, and
  troubleshooting. ([#422](https://github.com/huggingface/speech-to-speech/pull/422))

### Reliability and security

- Session teardown now preserves drain sentinels, releases capacity after setup
  failures, quarantines stuck pipeline units, reports them through `/v1/pool`,
  and prevents stale teardown signals from releasing the wrong session.
  ([#358](https://github.com/huggingface/speech-to-speech/pull/358))
- Speculative turn tracking no longer resurrects untracked revisions after a
  reset or LRU eviction, preventing reused turn IDs from suppressing responses.
  ([#391](https://github.com/huggingface/speech-to-speech/pull/391))
- Both MLX Audio Whisper and Lightning Whisper MLX now use the global MLX lock,
  preventing Metal command-buffer crashes when STT overlaps other MLX work on
  Apple Silicon. ([#379](https://github.com/huggingface/speech-to-speech/pull/379),
  [#387](https://github.com/huggingface/speech-to-speech/pull/387))
- Optional handler imports no longer reconfigure the host application's root
  logger, and setup/cleanup output now uses module loggers.
  ([#409](https://github.com/huggingface/speech-to-speech/pull/409))
- NLTK is upgraded to 3.10.0 to address CVE-2026-54293 / GHSA-p4gq-832x-fm9v.
  ([#400](https://github.com/huggingface/speech-to-speech/pull/400))

### Breaking change

The raw PCM WebSocket mode is now named `raw-websocket`. Commands using
`--mode websocket` must change to `--mode raw-websocket`; the old value is no
longer accepted. Realtime mode is unchanged and still supports both WebSocket
and WebRTC transports. ([#401](https://github.com/huggingface/speech-to-speech/pull/401))

### Packaging notes

- Smart Turn adds `huggingface-hub` and `onnxruntime` as standard dependencies.
- WebRTC support is available through `pip install "speech-to-speech[webrtc]"`.
- The Smart Turn model is downloaded from the Hugging Face Hub on first use
  unless a local model path is supplied.

### All merged changes

- Fix stuck pipeline units in session teardown by @A-Mahla in [#358](https://github.com/huggingface/speech-to-speech/pull/358)
- Add WebRTC transport for the OpenAI Realtime API by @A-Mahla in [#352](https://github.com/huggingface/speech-to-speech/pull/352)
- Bump the actions group with 2 updates by @dependabot in [#364](https://github.com/huggingface/speech-to-speech/pull/364)
- Add mic/speaker device selectors to the demo Settings UI by @dcolley in [#376](https://github.com/huggingface/speech-to-speech/pull/376)
- Acquire the global MLX lock in MLXAudioWhisperSTTHandler by @Hotragn in [#379](https://github.com/huggingface/speech-to-speech/pull/379)
- Add demo startup greeting and HF user attribution by @andimarafioti in [#389](https://github.com/huggingface/speech-to-speech/pull/389)
- Update star history chart by @andimarafioti in [#393](https://github.com/huggingface/speech-to-speech/pull/393)
- Add an LLM proxy for remote LLM backends by @A-Mahla in [#368](https://github.com/huggingface/speech-to-speech/pull/368)
- Add GitHub Trending achievement badge by @andimarafioti in [#398](https://github.com/huggingface/speech-to-speech/pull/398)
- Acquire the MLX lock in LightningWhisperSTTHandler by @varunsahni18 in [#387](https://github.com/huggingface/speech-to-speech/pull/387)
- Stop SpeculativeTurnTracker from committing untracked turns by @ztcools in [#391](https://github.com/huggingface/speech-to-speech/pull/391)
- Give every STT language an LLM language name by @Hotragn in [#395](https://github.com/huggingface/speech-to-speech/pull/395)
- Refine the GitHub Trending badge by @andimarafioti in [#399](https://github.com/huggingface/speech-to-speech/pull/399)
- Upgrade NLTK to 3.10.0 by @andimarafioti in [#400](https://github.com/huggingface/speech-to-speech/pull/400)
- Rename raw PCM mode to `raw-websocket` by @andimarafioti in [#401](https://github.com/huggingface/speech-to-speech/pull/401)
- Add user audio replay to the Realtime demo by @andimarafioti in [#390](https://github.com/huggingface/speech-to-speech/pull/390)
- Add missing Qwen3-TTS GGML features by @andimarafioti in [#403](https://github.com/huggingface/speech-to-speech/pull/403)
- Avoid configuring the root logger when importing optional handlers by @salignatmoandal in [#409](https://github.com/huggingface/speech-to-speech/pull/409)
- Add audio-input LLM STT bypass by @andimarafioti in [#298](https://github.com/huggingface/speech-to-speech/pull/298)
- Send `session.updated` after `session.update` by @hrqiang in [#413](https://github.com/huggingface/speech-to-speech/pull/413)
- Fix the `session.updated` event contract by @andimarafioti in [#417](https://github.com/huggingface/speech-to-speech/pull/417)
- Integrate the `session.updated` work by @andimarafioti in [#418](https://github.com/huggingface/speech-to-speech/pull/418)
- Refresh the star history chart by @andimarafioti in [#419](https://github.com/huggingface/speech-to-speech/pull/419)
- Document the Gemma 4 12B Realtime setup by @andimarafioti in [#422](https://github.com/huggingface/speech-to-speech/pull/422)
- Bump the actions group with 3 updates by @dependabot in [#406](https://github.com/huggingface/speech-to-speech/pull/406)
- Integrate Smart Turn v3.2 endpointing by @andimarafioti in [#192](https://github.com/huggingface/speech-to-speech/pull/192)

### New contributors

Thank you to @dcolley, @Hotragn, @varunsahni18, @ztcools,
@salignatmoandal, and @hrqiang for their first contributions to the project.

**Full changelog:** https://github.com/huggingface/speech-to-speech/compare/v0.2.11...v0.2.12