What Cartesia Sonic is
Cartesia Sonic is a text-to-speech model for real-time voice applications, offered through an API and the web. It produces speech with sub-90ms latency and supports 44 languages. Sonic can interpret emotional subtext and add non-verbal expressions, including laughter, to transcripts. Its localization feature carries emotion, tone, and speaker identity when adapting an audio clip into another language. Custom pronunciation dictionaries let you control how proper nouns and domain-specific terms are spoken. Managed Agents combine Ink-2 speech-to-text, a selected large language model, and Sonic-3.6 text-to-speech, with turn-taking and interruption handling. Agents can use Cartesia-provisioned numbers, imported Twilio numbers, or a carrier connected through SIP. Cartesia lists HIPAA, SOC 2 Type 2, GDPR, and PCI, and supports on-premise deployment, customer VPCs on AWS, GCP, or Azure, and OEM licensing. Plans include Free at no cost, Pro at $5 a month, Startup at $49 a month, Scale at $299 a month, and Enterprise with custom credits and agent usage. Commercial use is included on Pro and Sonic supports voice cloning.
Who it suits
Sonic suits developers building real-time voice applications through an API or web interface, especially when low latency, multilingual output, emotional delivery, or voice localization matters. Teams needing managed voice agents, SIP connectivity, compliance listings, or private deployment can use its broader deployment and agent options.
Good to know, and what to watch
- Sub-90ms latency for real-time voice applications
- Supports 44 languages
- Carries emotion, tone, and identity across localized audio
- Custom pronunciation dictionaries for names and domain terms
- On-premise and customer-VPC deployment options
- Managed Agents handle turn-taking and interruptions
- Free plan provides 20K monthly credits
- Free plan allows two concurrent Sonic requests
- Commercial use license starts with the Pro plan
- Enterprise uses custom credits and agent usage
Plans
Every plan and pricePro
Startup
Scale
Enterprise
Free plan
Getting started
- Open the Cartesia website
- Choose API or web access
- Create an account on the Free plan or select a paid plan
- Configure Sonic text-to-speech and pronunciation dictionaries
- Connect Managed Agents to Cartesia numbers, Twilio numbers, or SIP when needed
- Deploy on-premise, in an AWS, GCP, or Azure customer VPC, or through OEM licensing
Limits to check
The Free plan includes 20K credits per month, $1 prepaid agents per month, and two concurrent Sonic text-to-speech requests. Pro raises usage to 100K credits, $5 prepaid agents per month, and three concurrent requests; commercial use and instant voice cloning begin there.
The verdict
Pick Cartesia Sonic when your application needs fast, multilingual text-to-speech with emotional expression, localization, or managed voice agents. Its API, web access, deployment choices, and compliance listings also suit technical teams with varied infrastructure needs. Look elsewhere if the Free plan's credit and concurrency limits are too restrictive or if you need commercial use without moving to Pro.
Questions people ask
Is Cartesia Sonic free?
Yes. The Free plan costs nothing and includes 20K credits per month.
What does Pro cost?
Pro costs $5 a month, billed at $5/mo, with 100K monthly credits and a commercial use license.
Which platforms support Sonic?
Sonic is available through an API and the web.
Does Sonic support voice cloning?
Yes. Pro includes instant voice cloning, while Startup includes Pro voice cloning.
Can Sonic run in a private environment?
Cartesia says Sonic can run on-premise, in a customer VPC on AWS, GCP, or Azure, or through OEM licensing.
What do Managed Agents include?
Managed Agents combine Ink-2 speech-to-text, a chosen LLM, and Sonic-3.6 text-to-speech with turn-taking and interruption handling.
What the maker publishes
- API access
- Yes source
- Commercial use
- Yes source
- Concurrency
- The Free plan lists two concurrent Sonic text-to-speech requests, while Pro lists three. source
- Deployment
- Cartesia says Sonic can be deployed on-premise, in a customer VPC on AWS, GCP, or Azure, or through OEM licensing. source
- Emotion
- Sonic interprets emotional subtext and can insert non-verbal expressions such as laughter into transcripts. source
- Free plan
- Yes source
- Headquarters
- San Francisco, California, United States source
- Integrations
- Managed Agents can use Cartesia-provisioned numbers, import Twilio numbers, or connect a carrier over SIP. source
- Languages
- Sonic supports 44 languages. source
- Latency
- Sonic has sub-90ms latency. source
- Localization
- Sonic can localize an audio clip into other languages while carrying over emotion, tone, and speaker identity. source
- Platforms
- web,api source
- Product
- Sonic is a text-to-speech model for real-time voice applications. source
- Pronunciation
- Custom pronunciation dictionaries let users specify pronunciations for proper nouns and domain terms. source
- Security
- Cartesia lists HIPAA compliance, SOC 2 Type 2, GDPR, and PCI on the Sonic page. source
- Paid from
- 5 source
- Support
- Cartesia's documentation directs users to Discord to chat with the Cartesia team and community for help. source
- Voice agents
- Managed Agents combine Ink-2 speech-to-text, a chosen LLM, and Sonic-3.6 text-to-speech with turn-taking and interruption handling. source
- Voice cloning
- Yes source
- Voice cloning
- Instant voice cloning can use 10 seconds of audio, while Pro Voice Clone training uses 30 minutes or more from one speaker. source
Checked against the maker's pages on 1 Oct 2026.