Cartesia

HQ: United States
Founded Year: 2023
Company Type: Privately Held
Category: Speech AI
This is CXSpy's primary classification of the Cartesia software. The software may fit other categories as well.
Type: platform
CXSpy classifies CX software in one of two types (a platform or a tool). This is our view of the Cartesia software.

Description:Cartesia is a San Francisco voice AI company that builds real-time speech models used inside many of the AI agents now answering customer calls. It was founded in 2023 by Stanford researchers Karan Goel, Albert Gu, Arjun Desai, and Brandon Yang together with professor Chris Re, the group behind state space model research such as S4 and Mamba. That architecture lets Cartesia run speech synthesis with very low latency, which matters in a phone conversation where pauses sound unnatural. Its main products are Sonic, a text to speech model now on its third generation with support for 42 languages, Ink, a streaming speech to text model, and a managed agents platform for building complete voice agents. Models can run in Cartesia's cloud, on premises, or on device. The company raised a $64 million Series A led by Kleiner Perkins in March 2025 and a further $100 million in October 2025 with Kleiner Perkins, Index Ventures, Lightspeed, and NVIDIA. Customers listed by Cartesia include ServiceNow, Retell, Replicant, Forethought, Regal, and EliseAI.

Cartesia Speech AI Software

About Cartesia

Cartesia comes out of the Stanford AI lab, where its founders worked on state space models, an alternative to the transformer architecture that handles long sequences of audio efficiently. Karan Goel, Albert Gu, Arjun Desai, Brandon Yang, and Chris Re started the company in 2023 to apply that research to real-time voice.

The first product was Sonic, a text to speech model built to start speaking within a fraction of a second. Sonic 3, released in October 2025 alongside a $100 million funding round, supports 42 languages and quotes model latency of about 90 milliseconds. Cartesia later added Ink, a streaming speech to text model designed for conversational turns, and a managed agents platform that combines its models with telephony so customers can stand up a voice agent without stitching together several vendors.

For the contact center world, Cartesia matters mostly as an ingredient. Its customer page lists voice agent builders and CX software companies such as Retell, Replicant, Forethought, Regal, Maven AGI, Lorikeet, Thoughtly, and Bolna, along with ServiceNow. EliseAI, which automates property management communications, runs more than five million calls a month through Cartesia voices.

The company raised a $64 million Series A led by Kleiner Perkins in March 2025, then $100 million in October 2025 from Kleiner Perkins, Index Ventures, Lightspeed, and NVIDIA. Models can be deployed through Cartesia's API, inside a customer's own infrastructure, or on device.

Cartesia Features

1. Sonic text to speech with real-time streaming and support for 42 languages

2. Ink streaming speech to text built for conversational turn taking

3. Managed voice agents platform with built-in telephony

4. Voice cloning and a library of stock voices

5. State space model architecture designed for low latency

6. Deployment through cloud API, on premises, or on device

7. SDKs and integrations used by voice agent platforms

8. Enterprise security, volume pricing, and dedicated support

Cartesia Pricing

Cartesia publishes self-serve plans. The Free plan costs $0, Pro costs $5 per month, Startup costs $49 per month, and Scale costs $299 per month, with each tier including a larger allowance of credits for speech generation and transcription. Enterprise pricing is custom and adds volume discounts, on-premises deployment, and support commitments. For its managed agents product, Cartesia lists agent usage at $0.06 per minute and telephony at $0.014 per minute.