Home / Directory / AI SaaS tooling / Voice, speech & multimodal pipelines / Cartesia

Cartesia

Cartesia is a San Francisco voice AI company that builds its own real-time models, with Sonic for text to speech, Ink for speech to text, and Line for running phone and web voice agents.

7.9/10
Overall Score
Recommend

Cartesia is one of the fastest text-to-speech APIs a voice agent team can buy, with published credit pricing that starts free.

Best for

Developers and product teams building real-time voice agents for support, scheduling, sales, or healthcare calls, where the time before the first word of audio decides whether a caller stays on the line.

Not ideal for

Audiobook, podcast, and character-voice production, where long passages and expressive narration matter more than speed, and buyers who want a very large library of stock voices.

Verdict

Cartesia's Sonic models are among the fastest voices on the market, and Sonic-3.5 led the Artificial Analysis Speech Arena in May 2026. The company grew out of Stanford research on state space models, a model design that keeps latency low, and it now sells speech to text and a managed voice agent platform alongside the voices. More than 50,000 customers use it, and it has raised $191 million. Quality on long narration still trails ElevenLabs, but for a voice agent that has to answer in real time, Cartesia is a strong first choice.

Score Breakdown

These scores cover how natural Cartesia's voices sound, how fast they respond in a live call, how many languages and voices it offers, and how clear its pricing is, and then how it does as a company on innovation, communication, pricing, and contract terms.

Buyer outcomes

Voice quality and naturalness
8.2
Latency for real-time agents
8.6
Language and voice coverage
7.6
Pricing transparency
8.0

Company & commercial

Innovation & product leadership
8.6
Project management & communication
7.6
Pricing
7.4
Contract fairness
7.2

Who Uses Cartesia, Where It Fits on Price, and Its Competitors

Cartesia sells mostly to engineering teams that build voice agents, including voice agent platforms such as Retell AI and enterprises such as ServiceNow, Cresta, and Quora. Most start with the Sonic text-to-speech API, then add Ink for transcription or move whole agents onto Line, which handles phone numbers, calls, and evaluations. The company was founded in 2023 by Karan Goel, Albert Gu, Arjun Desai, Brandon Yang, and Chris Ré, who met through Stanford's AI lab, and it has about 170 people.

On price, Cartesia sits in the middle of the market. Plans run from free to $299 a month, with overage rates from $38 to $65 per million credits, and enterprise contracts are quoted. Its closest competitors are ElevenLabs, Deepgram, Rime, and AssemblyAI.

Pricing

Cartesia publishes monthly plans that bundle model credits with prepaid voice agent dollars. Prices are as of November 2026.

PlanPriceWhat stands out
Free$020,000 credits and $1 of voice agent use a month.
Pro$5 a month100,000 credits, a commercial license, and instant voice cloning.
Startup$49 a month1.25 million credits, $49 of agent use, and two professional voice clones.
Scale$299 a month8 million credits, SSO, DPAs and BAAs, and a shared Slack channel.
EnterpriseQuotedCustom concurrency and on-premise or private cloud deployment.

The Field at a Glance

How Cartesia compares with other voice, speech, and multimodal vendors we reviewed, by Overall Score and relative typical engagement cost.

6 7 8 9 Overall Score $ $$ $$$ $$$$ Relative typical engagement cost ElevenLabs Deepgram Gladia AssemblyAI Deepslate Speechmatics Rime Cartesia 7.9
Cartesia Deepgram AssemblyAI Gladia Rime ElevenLabs Deepslate Speechmatics

Cartesia scores 7.9 in voice, speech, and multimodal pipelines, behind Deepgram (8.0), level with AssemblyAI (7.9), Gladia (7.9), and Rime (7.9), and ahead of ElevenLabs (7.8), Deepslate (6.9), and Speechmatics (6.8).

Compared with …

How Well Cartesia Fits Each Voice AI Use Case

We rated Cartesia against the jobs teams bring to a speech model provider, drawing on its pricing page, independent latency tests, and benchmark write-ups.

Use caseFitNotes
Real-time phone and web agentsStrongSonic streams first audio in under 100 milliseconds over WebSocket in independent tests.
Multilingual customer supportStrongSonic-3 covers 42 languages, including nine Indian languages.
Transcription for live callsStrongInk is built for noisy phone audio and bills per second.
Managed voice agentsMixedLine is newer than dedicated agent platforms such as Vapi and Retell.
Audiobooks and narrationWeakLong passages can drift, and expressive narration trails ElevenLabs.
Large stock voice librariesMixedThe catalog is smaller than ElevenLabs, though cloning fills gaps.

Review Excerpts

These excerpts come from a developer comment on Hacker News and from buyer concerns we found in Cartesia's pricing details and independent benchmark write-ups. Paid reviews and pay-for-play sites were excluded.

What people like

“Check out Sonic (cartesia.ai). Great quality, very fast - but with a few kinks to work out (going off the rails on long utterances, random sounds, etc).”

nmfisher, in a comment on Hacker News
How it's used

“We put Sonic behind our scheduling agent because callers stopped talking over it. The first word comes back fast enough that the pause feels normal.”

Representative buyer view
What people don't like

“Credits made budgeting harder than it should be. We had to work out characters per minute, then add agent minutes, telephony, and the LLM on top.”

Representative buyer view

Where customer reviews couldn't be found, our analysts attempted to give a voice to buyer sentiment, drawing on real information from review sites, forums, and product documentation.

Company Facts

Founders
Karan Goel (CEO), Albert Gu, Arjun Desai, Brandon Yang, and Chris Ré
Founded
2023
Headquarters
San Francisco, California
Funding
$191 million in total, including a $100 million Series B led by Kleiner Perkins in October 2025
Customers
More than 50,000 as of March 2026
Website
cartesia.ai

Key Leadership

Karan Goel

Karan Goel

Co-founder and CEO

He earned a master's degree at Carnegie Mellon and a PhD at Stanford, where he met his co-founders.

Albert Gu

Co-founder

He met Goel at Stanford, where the group's research on state space models became the basis for Cartesia.

Methodology

We scored Cartesia on eight criteria, from voice quality and naturalness to contract fairness, and the Overall Score is their simple average on a 0 to 10 scale. The scores draw on reviews from public review sites and forums, with paid reviews and pay-for-play sites excluded, plus a hands-on look at the product, its pricing, and its documentation and training. Scores can change as we add reviews and product checks.

← Back to Voice, speech & multimodal pipelines

Sources

Facts on this page

  • He earned a master's degree at Carnegie Mellon and a PhD at Stanford, where he met his co-founders: Dell Technologies Capital
  • He met Goel at Stanford, where the group's research on state space models became the basis for Cartesia: Dell Technologies Capital

Company facts

Review excerpts

Quote sources

  • “Check out Sonic (cartesia.ai). Great quality, very fast -…” (nmfisher, in a comment on Hacker News): Hacker News