
by Cartesia
Cartesia is an AI lab that offers real-time text-to-speech, speech-to-text and a platform for building voice agents, all through an API.
APIUsage-based
Step 1: Try the models in the browser playground and get an API key.
Step 2: Install the Python or JavaScript client library.
Step 3: Open a WebSocket connection and send text in chunks with a voice ID.
Step 4: Play the returned audio chunks, for example with ffplay.
Step 5: Stream audio to the speech-to-text model to get partial and final transcripts.
Use ephemeral access tokens so your API key stays private. The docs link a guide for authenticating client applications.
You push text chunks into a WebSocket context and receive audio chunks back. The system keeps context history, so speech stays continuous in tone and prosody.
If overages are enabled for the workspace, requests continue and extra usage is billed on top of the subscription. If not, requests fail until renewal or an upgrade.
No. Credits are only used by successful requests.
Cartesia VoiceCartesia provides real-time voice AI through one API: text-to-speech, speech-to-text and managed voice agents for phone calls. Speech is supported in 44 languages.
Cartesia SonicCartesia Sonic applies AI to low-latency generative voice model and API for realtime conversational speech and expressive text-to-speech.
SonioxSoniox provides speech APIs for transcription, text to speech and real-time speech translation. It supports 60+ languages through REST and WebSocket interfaces.
OpenAI Speech to TextOpenAI Speech to Text applies AI to AI speech-to-text API for transcribing and translating audio with OpenAI speech models.