Cartesia — AI for Music, Audio & Voice Production

Cartesia is a voice platform aimed at developers building real-time voice products. The homepage lists Sonic for text-to-speech, Ink for speech-to-text, Managed Agents for voice agents on phone calls, and a voice library. It says the models are built on state space models, which the company credits for low latency, and it offers cloud, on-premise and on-device deployment, in-region inference and an API, docs, cookbooks and an MCP server. The page gives no latency numbers. It lists speech in 44 languages and cites Artificial Analysis leaderboard rankings as a vendor claim. The pricing page is clear. Free is $0 with 20K monthly credits, about 27 text-to-speech minutes, and covers text-to-speech and speech-to-text. Pro at $5 per month adds a commercial use license and instant voice cloning. Startup at $49 adds professional voice cloning and organizations, Scale at $299 adds priority support, and Enterprise is custom. Text-to-speech costs 1 credit per character and speech-to-text 3 credits per second. Voice agents are $0.03 per minute, telephony is $0.014 per minute, and LLM usage during calls is billed separately. Free has only 2 concurrent TTS requests and no professional voice clones. Cartesia fits teams building voice apps and agents; if you only want a simple narration studio, a tool with a visual editor may be easier.