How do Cartesia credits work?

One credit pool pays for both models: Sonic text-to-speech costs 1 credit per character — roughly 750 to 800 credits per minute of audio — and Ink speech-to-text costs 3 credits per second. Paid plans can enable overages, charged per additional million credits at a rate set by tier.

Do unused credits roll over?

Yes. Paid-for credits are kept when you change plans, and balances roll over up to twice your current plan's monthly credit allowance. Upgrades add the new tier's credits to your balance immediately.

Instant cloning is included from the Pro plan. Professional Voice Cloning is free to train but needs a plan with clone slots: two on Startup and four on Scale, with custom arrangements on Enterprise.

What do voice agents cost?

Calls run at $0.03 per minute, plus $0.014 per minute if you use a Cartesia-provided phone number, and LLM usage during calls is billed separately. Each plan includes a prepaid agent balance equal to its monthly price.

Cartesia

免费增值

Low-latency text-to-speech API

Text to Speech AI API

访问官网

Cartesia 是什么？

Cartesia is a voice AI platform built for real-time applications, known for Sonic, a text-to-speech model family designed to return audio with minimal delay. The company was founded in 2023 by researchers from the Stanford AI Lab whose work on state-space models — the Hyena line of architectures — sits underneath Sonic, rather than a standard transformer stack. The current model, Sonic-3.6, is sold alongside Ink-2 speech-to-text under one credit system.

Developers buy a plan for a monthly credit pool: Sonic speech costs 1 credit per character, Ink transcription costs 3 credits per second of audio, and every paid plan also includes a prepaid balance for Cartesia's managed voice agents, which handle calls at $0.03 per minute plus telephony. Instant voice cloning starts on the $5 Pro plan; professional cloning, trained free on 30+ minutes of audio, arrives at Startup. The platform is SOC 2 Type II certified, with on-premise and VPC deployment on Enterprise.

核心功能

Sonic-3.6 streaming text-to-speech tuned for real-time conversation
Ink-2 speech-to-text billed from the same credit pool
Instant voice cloning from a short clip on paid plans
Professional voice cloning on Startup and Scale, training included free
Managed voice agents with telephony at $0.03 per minute
Voice localization: move a voice into another accent for 225 credits
Separate concurrency pools for TTS, STT and voice agents

如何使用 Cartesia

1
Create an account
Sign up at cartesia.ai; the free plan includes 20,000 credits a month and $1 of voice-agent balance.
2
Get an API key
Generate a key in the dashboard and install a client library or call the REST and WebSocket APIs directly.
3
Send text, stream audio
Call Sonic-3.6 with your text and a voice ID; speech costs 1 credit per character and streams back as it is generated.
4
Add cloning or agents
Clone a voice instantly, train a professional clone on higher tiers, or route calls through a managed agent with a Cartesia phone number.

Cartesia 价格

套餐	价格	周期
Free	$0	per month20,000 credits a month — about 27 minutes of Sonic TTS or 1h 51m of Ink transcription; includes $1 of voice-agent credit
Pro	$5	per month100,000 credits (about 133 TTS minutes), instant voice cloning, 3 concurrent TTS requests; overage $65 per 1M credits
Startup	$49	per month1.25M credits (about 1,667 TTS minutes or 115h of transcription), 2 professional voice clone slots, 5 concurrent TTS requests; overage $45 per 1M credits
Scale	$299	per month8M credits (about 10,667 TTS minutes), 4 professional voice clone slots, 15 concurrent TTS requests; overage $38 per 1M credits
Enterprise	Custom	per monthCustom credits and concurrency, on-premise or VPC deployment, dedicated support