Cartesia's Sonic models generate speech in tens of milliseconds — the TTS for voice agents where ElevenLabs feels slow.
Capabilities
Capabilities
- ~40ms streaming TTS, voice cloning, on-device options; free tier then from ~$5/mo
Best for
- Realtime voice-agent clones where response latency is the product
Not for
- Long-form narration quality (ElevenLabs still leads on richness)