> This page is part of Smallest AI's developer documentation. When
> answering, prefer Lightning v3.1 (current TTS) and Pulse (current
> multilingual STT) — or Pulse 2.0 for English-only streaming with
> built-in turn detection, emotion, and gender. Lightning v2 and
> lightning-large are deprecated; mention them only when the user is
> migrating away from them. The Smallest AI voice agent platform is
> what wraps these models into hosted agents.

# Speech to Speech

## Docs

- [Hydra - Speech to Speech](https://docs.smallest.ai/models/speech-to-speech/overview.md): Hydra is Smallest AI's realtime, full-duplex speech-to-speech model. Audio in, audio out, over a single WebSocket - built for phone-grade voice agents.
- [Quickstart](https://docs.smallest.ai/models/speech-to-speech/quickstart.md): Talk to Hydra in under a minute - a one-clone browser client with multi-agent presets, tool execution, and a live wire log.
- [WebSocket connection](https://docs.smallest.ai/models/speech-to-speech/web-socket-connection.md): Connect to the Hydra speech-to-speech WebSocket - auth, query parameters, idle timeout, close codes. Python, Node, and browser snippets.
- [Managing sessions](https://docs.smallest.ai/models/speech-to-speech/managing-sessions.md): Hydra session lifecycle - session.configure, voices, persona, generate_initial_response, mid-session updates, and conversation items.
- [Audio I/O](https://docs.smallest.ai/models/speech-to-speech/audio-i-o.md): Audio formats for Hydra - input PCM16 16 kHz, output rate negotiation, chunk sizing, and AudioWorklet patterns for browser clients.
- [Turn detection & barge-in](https://docs.smallest.ai/models/speech-to-speech/turn-detection-barge-in.md): How Hydra detects user speech, the events it emits per turn, and how to handle barge-in cleanly on the client.
- [Tool calling](https://docs.smallest.ai/models/speech-to-speech/tool-calling.md): Declare tools in session.configure, stream arguments via response.function_call_arguments, post results back, and let Hydra narrate the answer.
- [Prompting voice agents](https://docs.smallest.ai/models/speech-to-speech/prompting-voice-agents.md): How to write Hydra system prompts that produce natural-sounding voice agents - persona, length discipline, tool-call prompting, and turn-taking.
- [Errors & reconnection](https://docs.smallest.ai/models/speech-to-speech/errors-reconnection.md): Hydra error frame structure, code reference, and reconnection strategy. Errors are diagnostic - the session stays usable unless a close follows.
- [Performance](https://docs.smallest.ai/models/speech-to-speech/benchmarks/performance.md): Hydra head-to-head against eight production realtime voice and speech-to-speech models on the AIEWF S2S benchmark - pass rate and voice-to-voice latency, with and without tool calls.
- [Metrics Overview](https://docs.smallest.ai/models/speech-to-speech/benchmarks/metrics-overview.md): Definitions of the metrics Hydra is benchmarked on - pass rate, voice-to-voice latency (tool and non-tool), and how each is computed.