Skip to main content
Chariot offers three ways to synthesize speech. They produce the same audio from the same voices. The difference is latency and how you deliver text.

​
Decision guide

REST (sync)

You have short, complete text and want one finished WAV file. Simplest to integrate.

HTTP streaming

You have one complete utterance but want playback to start as early as possible.

WebSocket

Text arrives incrementally (e.g. from an LLM), or you need a long, low-latency, interactive session.

Multi-context WebSocket (beta)

A voice agent that must stop speech mid-sentence when the user interrupts, or run more than one reply at a time, on one connection.

​
Comparison

​
Rules of thumb

  • Start with REST. If you’re generating short clips and don’t need sub-second first audio, POST /v1/tts is the least code.
  • Reach for HTTP streaming when you already have the full sentence but want the user to hear audio sooner, for example reading a chat reply aloud.
  • Use the WebSocket when text is produced token-by-token (streaming from an LLM), when you’re driving a voice agent, or when your script exceeds a single request’s model limit. You send input.text messages as text becomes available, then input.flush to generate audio before the buffer reaches the model limit. See WebSocket streaming.

​
Which WebSocket?

Both WebSockets stream incrementally with the same voices, latency, and billing. Choose by what your application needs to do with speech once it has started.
  • Start with /v1/tts/ws if you only ever play one stream of speech and never need to stop it early. It is the simpler protocol.
  • Use /v1/tts/ws-multi for voice agents. When the user interrupts, cancel the reply that is speaking and start the next one on the same connection, without reconnecting. See Multi-context WebSocket.
Concurrency limits apply to sync, HTTP streaming, and active WebSocket generations. HTTP returns 429 when the limit is reached; the WebSocket closes with 4429; the multi-context WebSocket ends only the affected context. If you need many simultaneous generations, see Rate limits & concurrency before choosing a transport.