Cartesia
Real-time voice AI APIs: Sonic text-to-speech and Ink speech-to-text built on state space models, with voice cloning and voice agents
Overview
Cartesia bills itself as "#1-ranked for realtime voice". It sells Sonic, a text-to-speech model family (Sonic-3.6, August 2026), and Ink, a speech-to-text family (Ink-2, July 2026), both built on state space models rather than transformers. Cartesia claims sub-90 ms text-to-speech and 100 ms transcript latency with built-in turn detection, speech in 44 languages, instant and professional voice cloning, and a hosted voice agent product billed per minute. CEO Karan Goel and co-founder Albert Gu (co-author of Mamba) lead the company; it raised a $27M seed led by Index Ventures in December 2024 and a $64M Series A led by Kleiner Perkins in March 2025. For a voice-quality-first rival see ElevenLabs.
The Verdict
Who Should Use Cartesia?
Best For
- Voice agent builders who need the lowest possible speech latency
- Developers who want TTS and STT from one vendor
- Products that need cloned voices in many languages
Not Ideal For
- Audiobook or film work where realism matters more than speed
- Buyers who want many third-party reviews before choosing
What's Great
- Very low latency, a recurring point in comparisons
- Cheap entry plans and a free tier
- One credit system for TTS, STT and cloning
Watch Out For
- ElevenLabs is usually rated more natural and stronger at cloning
- Latency figures are vendor claims; end-to-end times are higher
- Few independent user reviews so far
Pricing
View all features & details
Models
- Sonic text-to-speech (Sonic-3.6)
- Ink speech-to-text (Ink-2) with keyterms and turn detection
- Instant and professional voice cloning
- Multilingual voices
Billing
- TTS: 1 credit per character
- STT: 3 credits per second of audio
- Voice agents: $0.03 per minute plus telephony
How It Compares
| Feature | Cartesia | ElevenLabs | AssemblyAI |
|---|---|---|---|
| Main strength | Lowest-latency voice | Voice realism and cloning | Transcription accuracy |
| Text-to-speech | Yes (Sonic) | Yes | No |
| Speech-to-text | Yes (Ink) | Yes | Yes |
| Entry paid plan | $5/mo | See vendor | Pay as you go |
Sources: Cartesia · Cartesia blog · Maginative (Mar 2025) · AssemblyAI pricing
Is this your tool? Add a badge