Build a voice agent

Choose a complete voice-agent tutorial, a direct API starter, or a supported voice platform.

Start with the LiveKit tutorial for a complete, working voice agent. Use a direct API starter when you already own transport, transcription, turn detection, and orchestration. Rime supplies text-to-speech in either case. Rime speech-to-text can also handle transcription: it streams partial transcripts while the user speaks, and your agent sends the final one to its language model. To replace the whole pipeline with one API, use Prism, Rime’s speech-to-speech API.

Keep RIME_API_KEY on the server. Browser clients should call your application server or connect through a server-side WebSocket bridge rather than authenticate with Rime directly.

​

Choose an implementation

LiveKit voice agent

Run an end-to-end browser agent with LiveKit transport, turn detection, speech recognition, agent logic, and Rime speech.

Pipecat with Prism

Run a browser voice agent with Prism’s model and turn handling through the Pipecat preview.

Direct API starters

Add a server-side Rime speech route and a browser voice interface to your application stack.

• Next.js
• Vite + React
• Express
• Node.js
• FastAPI

Voice platforms

Use Rime through a framework or service that supplies transport, telephony, or agent orchestration.

• LiveKit
• Pipecat
• Vapi
• Daily

WebSocket API

Build directly on persistent synthesis connections when you need incremental text, word timestamps, context management, or interruption control.

Prism speech-to-speech

Send the caller’s audio to Prism over one WebSocket. Prism takes turns, handles interruptions, and replies in speech with your instructions and tools.

​

Choose Prism or a pipeline

Prism and a pipeline solve the same problem in different ways. A pipeline gives you a language model of your choice, more languages, and the full voice catalog. Prism gives you turn-taking, interruptions, and fillers without building them yourself. Here’s how the two compare:

	Prism	Speech-to-text, your LLM, and Rime text-to-speech
Language model	Prism’s own model, with your instructions and tools	Any model that you choose
Turn detection	Prism	Your application, framework, or turn detector
Languages	English	Your speech-to-text decides, and Rime’s models for speech
Voices	astra, aurora, brando, wawona	The Coda and Mist v3 catalogs, with pronunciation and speed controls
Reply text	Streams as Prism writes it. Your application cannot change it before Prism speaks it.	Your application can check or change it before text-to-speech
Your application handles	Audio capture and playback, playback reports, and tools	Audio, turn detection, the LLM, tools, and orchestration
Start here	Prism SDK quickstart	LiveKit tutorial

​

Rime is the speech layer

Layer	Responsibility
Audio transport	Carries microphone input and synthesized audio between the user and the agent.
Speech recognition	Converts the user’s audio into text for the agent, with Rime speech-to-text or another provider.
Agent logic	Uses application rules or a language model to decide what to say.
Speech generation	Sends the response text to Rime and returns natural audio to the user.

The LiveKit tutorial implements the complete path. The direct API pages are application starters: they demonstrate the Rime proxy and browser interface while leaving the production speech-recognition and agent-logic choices to your application.

​

Design for live conversation

Stream speech

Start playback while the response is still being generated and choose an audio format for your transport.

Handle interruptions

Use timestamps, context IDs, segmentation, and clear operations to coordinate turns and barge-in.

Reduce latency

Measure time to first audio and tune the synthesis path for responsive conversations.

Choose a voice

Match a voice, model, and language to the agent’s audience and role.