# Neuphonic Documentation > Developer documentation for the Neuphonic API, powered by NeuTTS-Max, our cloud text-to-speech model: low-latency text-to-speech over SSE and WebSocket streaming, voice cloning, 44 languages, word timings, pronunciation control and voice agents, with Python and JavaScript SDKs. Notes: - Authentication: generate an API key in the Playground (https://app.neuphonic.com/) and send it as the `X-API-KEY` header (SSE/REST) or the `api_key` query parameter (WebSocket). - Endpoints: SSE at `POST https://api.neuphonic.com/sse/speak/{lang_code}`; WebSocket at `wss://api.neuphonic.com/speak/{lang_code}`. On the WebSocket, end each piece of text with a `` token. - Model: the Neuphonic cloud API runs on NeuTTS-Max, and every voice uses it. Choose a voice with `voice_id`; list voices, and the languages each supports, with `client.voices.list()` or `GET /voices`. The chosen voice must be available for the chosen language. - Other TTS parameters: `lang_code`, `speed` (snaps to 0.7, 1.0 or 1.5), `sampling_rate` (8000, 16000, 22050 or 24000; default 24000), `encoding` (`pcm_linear` or `pcm_mulaw`), `timestamps` and `ipa_key`. - Word timings: set `timestamps=true` on the WebSocket to receive an `alignment` object with per-word start times, durations and character offsets. - Pronunciation: wrap IPA in a `` tag inline, available in ten languages (English, German, Dutch, French, Italian, Spanish, Portuguese, Russian, Hindi and Korean). Pauses: `` or SSML ``. - SDKs: Python `pip install pyneuphonic` (Python 3.9+); JavaScript `npm install @neuphonic/neuphonic-js` (Node.js 18+). ## Getting started - [Quickstart](https://docs.neuphonic.com/): Get an API key, install the Python or JavaScript SDK, and run SSE, WebSocket and agent examples ## Text-to-speech - [How to use text-to-speech](https://docs.neuphonic.com/build-group/text-to-speech): Generate speech over SSE or WebSocket streaming with cURL, wscat, Python and JavaScript - [Speech synthesis: SSE and WebSockets](https://docs.neuphonic.com/build-group/text-to-speech#speech-synthesis): When to use single-request SSE versus WebSocket streaming, and the `` token - [Text-to-speech configuration](https://docs.neuphonic.com/build-group/text-to-speech#text-to-speech-configuration): All TTS parameters, including voice selection via `voice_id`, speed, sampling rate and encoding - [Return configuration](https://docs.neuphonic.com/build-group/text-to-speech#return-configuration): WebSocket response format and the 60-second idle timeout - [Word timings](https://docs.neuphonic.com/build-group/text-to-speech#word-timings): `timestamps=true` and the `alignment` object for highlighting text as it is spoken - [Pronunciation](https://docs.neuphonic.com/build-group/text-to-speech#pronunciation): Correct a word's pronunciation with IPA in `` tags, in ten languages - [Inline pauses](https://docs.neuphonic.com/build-group/text-to-speech#inline-pauses): Insert silence with `` or SSML `` tags - [Providing a context ID](https://docs.neuphonic.com/build-group/text-to-speech#providing-a-context-id): Tag audio chunks with an ID to track or interrupt a request - [Resetting the stream](https://docs.neuphonic.com/build-group/text-to-speech#resetting-the-stream): Send `` to discard buffered text and start a fresh utterance without closing the WebSocket, e.g. on barge-in - [Languages](https://docs.neuphonic.com/build-group/languages): All 44 languages supported by NeuTTS-Max, their `lang_code` values and available accents ## Voices - [How to clone a voice](https://docs.neuphonic.com/build-group/voice-cloning): Prepare a clean sample, create a voice with `POST /voices`, list voices and delete cloned voices ## Agents - [Using the agents WebSocket API](https://docs.neuphonic.com/how-to-use-agents/): Create, configure, list and run voice agents with custom prompts and greetings - [Connecting MCP servers](https://docs.neuphonic.com/how-to-use-agents/connecting-mcp-servers): Give an agent tools by connecting your own MCP server - [Using LiveKit WebRTC](https://docs.neuphonic.com/how-to-use-agents/using-webrtc): Request a token and connect to a Neuphonic agent with a LiveKit WebRTC client ## Integrations - [LiveKit](https://docs.neuphonic.com/integrations/livekit): Use Neuphonic TTS in LiveKit real-time voice agents - [Pipecat](https://docs.neuphonic.com/integrations/pipecat): Use Neuphonic TTS in Pipecat voice and video agents - [Twilio](https://docs.neuphonic.com/integrations/twilio): Connect Neuphonic agents to Twilio phone calls, including inbound call setup - [VideoSDK](https://docs.neuphonic.com/integrations/video-sdk): Use Neuphonic TTS with VideoSDK voice and video agents ## Reference - [API reference](https://api.neuphonic.com/docs): Interactive reference for all REST endpoints - [OpenAPI specification](https://api.neuphonic.com/openapi.json): Machine-readable OpenAPI schema for the API - [Changelog](https://docs.neuphonic.com/changelog): API updates, including word timings and pronunciation control (September 2026) ## Optional - [Python SDK source](https://github.com/neuphonic/pyneuphonic): pyneuphonic repository, including the TTSConfig and AgentConfig models - [JavaScript SDK source](https://github.com/neuphonic/neuphonic-js): neuphonic-js repository and examples - [Open NeuTTS models](https://github.com/neuphonic/neutts): Run the open on-device NeuTTS models locally - [Playground](https://app.neuphonic.com/): Try voices, generate API keys and look up IPA with the pronunciation tool - [Neuphonic website](https://www.neuphonic.com/): Product, model and deployment overview - [Build category](https://docs.neuphonic.com/category/build): Index of build guides - [Agents category](https://docs.neuphonic.com/category/how-to-use-agents): Index of agent guides - [Integrations category](https://docs.neuphonic.com/category/integrations): Index of integration guides