Cartesia Pricing (2026): Every Plan, Credit & Overage Rate

Cartesia pricing starts at $0 with 20K credits a month, and the paid plans cost $5 (Pro), $49 (Startup), and $299 (Scale) a month before custom Enterprise pricing.

One credit pool pays for Sonic text-to-speech at about 1 credit per character and Ink-2 speech-to-text at 3 credits per second, while Managed Agents bill $0.06 a minute.

I worked through the full rate card and its billing FAQ. Cartesia pricing is harder to read than the tier list suggests, because one credit pool pays for both speech generation and transcription while voice agents draw from a separate dollar balance.

After running the numbers, I'd put an early production agent on Startup. Scale becomes worth its price once you need more than 20 concurrent calls or about 6.8M credits a month, the point where Startup plus overage passes $299.

To show where that line falls, I price each plan below and total up a typical month, overages included.

TL;DR: Cartesia Pricing

• Cartesia pricing has four self-serve tiers. Free is $0, Pro is $5, Startup is $49, and Scale is $299 per month. Enterprise is priced by sales.

• One credit pool covers TTS and STT. Sonic-3.6 speech costs about 1 credit per character, and Ink-2 transcription costs 3 credits per second of audio.

• Voice agents bill in dollars, at $0.06 per minute plus $0.014 per minute on a Cartesia phone number, before any LLM usage. Each paid plan adds prepaid agent dollars equal to its price.

• Overages are optional on paid plans and cost $65, $45, or $38 per 1M credits, depending on tier. With overages off, requests fail once the balance hits zero.

• Seats are unlimited on every plan, and concurrency is each plan's main limit, from 2 TTS requests on Free to 15 on Scale.

Cartesia Pricing Plans: At a Glance

Every plan comes with two allowances, a monthly pool of model credits and a prepaid dollar balance for Managed Agents.

Cartesia pricing verified against the vendor's pricing page on October 8, 2026. Verify with the vendor before you commit.

How Cartesia Pricing Works: Credits and Agent Dollars

Cartesia pricing runs on two balances: Sonic (TTS) and Ink (STT) use model credits, while Managed Agents use the prepaid dollar balance, billed per minute.

The credit pool itself is shared across both models. On Scale, the 8M monthly credits go to STT, TTS, or any mix of the two, so an hour of Ink-2 transcription (10,800 credits) uses roughly as many credits as 14 minutes of Sonic speech.

Four billing rules determine how long your credits last:

• Unused credits roll over until they reach 2x the monthly allowance.

• An upgrade adds the new tier's credits to your current balance right away.

• A downgrade or cancellation takes effect at the end of the billing period, and you keep the credits you already paid for.

• Failed requests cost nothing, since Cartesia's docs say errors don't consume credits.

Cartesia Pricing by Plan: What Each Tier Includes

Free: $0/month

What's included: 20K credits a month, which Cartesia converts to about 27 minutes of Sonic-3.6 audio or 1 hour 51 minutes of Ink-2 transcription.

The $1 agent balance buys about 17 minutes of Managed Agent calls. You also get 1 Cartesia phone number, 2 concurrent TTS requests, 8 concurrent STT requests, and 8 concurrent calls.

Best for: Developers adding Sonic or Ink to a prototype to hear the output and check latency before paying.

Pros

✅ Both TTS and STT run on the free tier. You can test speech generation and transcription together at $0.

✅ A working phone number is included. You can call a test agent from your phone without upgrading.

Cons

❌ Commercial use requires Pro. The Free plan carries no commercial use license.

Pro: $5/month

What's included: 100K credits a month, or about 133 minutes of speech or 9 hours 16 minutes of transcription, plus a $5 agent balance worth about 1 hour 23 minutes of calls.

Pro includes the commercial use license and Instant voice cloning. Concurrency caps at 3 TTS requests, 12 STT requests, and 12 calls, with 3 phone numbers.

Best for: Independent developers and side projects launching a small commercial voice feature.

Pros

✅ The cheapest commercial license on Cartesia's rate card, at $5 a month.

✅ You can clone a voice from 10 seconds of audio, per Cartesia, and cloned-voice speech bills at the standard rate of about 1 credit per character.

✅ Credits roll over up to 200K, so a slow month's unused credits carry into the next one.

Cons

❌ Running over costs more here than on any other tier. Pro overages bill $65 per 1M credits, against $45 on Startup and $38 on Scale.

❌ Three concurrent TTS requests can limit an agent on your own stack once more than 3 callers are waiting on a spoken reply at once.

Startup: $49/month

What's included: 1.25M credits a month, or about 1,667 minutes of speech or 115 hours 44 minutes of transcription. The $49 agent balance covers about 13 hours 37 minutes of calls.

Startup adds Pro voice cloning with 2 slots per organization and the Organizations feature. Each Professional Voice Clone needs 30+ minutes of your audio. Concurrency rises to 5 TTS requests, 20 STT requests, and 20 calls, with 5 phone numbers.

Best for: Seed and Series A teams running an early production agent with a branded voice.

Pros

✅ Overage drops to $45 per 1M credits, 31% below Pro's rate.

✅ Both Professional Voice Clone slots train for free, per Cartesia's FAQ.

Cons

❌ TTS concurrency is one-quarter of STT's. Startup allows 5 simultaneous TTS requests and 20 for STT, so teams sending Sonic traffic from their own orchestration are likely to hit that limit first.

❌ Priority support starts at Scale. Startup's plan card doesn't list it.

Scale: $299/month

What's included: 8M credits a month, or about 10,667 minutes of speech or 740 hours 44 minutes of transcription, plus a $299 agent balance worth about 83 hours of calls.

Scale adds priority support and 4 Professional Voice Clone slots. Concurrency goes to 15 TTS requests, 60 STT requests, and 60 calls, with 10 phone numbers.

If you're a startup, Cartesia's Startups Grant gives accepted teams 12 months on Scale at no cost, with 8M credits and $299 in agent credit each month. Cartesia aims to respond within a week, and when the grant ends, the account moves to $299 a month unless you switch plans.

Best for: Production voice apps that run more than 20 concurrent calls at peak.

Pros

✅ $38 per 1M credits is the lowest overage rate on the card, which matters once usage runs past 8M a month.

✅ 60 concurrent calls and 60 STT streams, 3x what Startup allows.

Cons

❌ Cartesia offers no tier between $49 and $299. When you outgrow Startup's 1.25M credits, your options are to either move to a tier that costs 6.1x more or pay overage on Startup.

❌ A signed BAA requires an Enterprise contract. Cartesia lists DPAs and BAAs on that tier only.

Enterprise: Custom Pricing

What's included: Custom credit and concurrency limits, plus volume pricing. Enterprise also carries DPAs and BAAs, SSO, a shared Slack channel, and security questionnaire support.

Cartesia's docs list Zero Data Retention and regional endpoints as Enterprise-only, and on-prem or VPC deployment is arranged with sales.

Best for: Healthcare, finance, and government teams that need a signed BAA or data residency before audio touches the API.

Pros

✅ Enterprise is the only tier that lists a BAA. Under HIPAA, a covered entity needs a written business associate agreement before a vendor creates, receives, maintains, or transmits PHI on its behalf (45 CFR 164.504(e)). Zero Data Retention is Enterprise-only too.

✅ Deployment on-prem or in your own VPC is available through sales.

Cons

❌ Every Enterprise quote goes through sales. Cartesia publishes no Enterprise rate, and the self-serve overage prices don't indicate what volume pricing will cost.

Cartesia TTS Pricing

Cartesia TTS pricing is about 1 credit per character on Sonic-3.6. It's approximate because transcript pre-processing can change the character count slightly. Cartesia's FAQ puts one minute of generated audio at 750 to 800 credits.

All minute figures below use the 750-credit end of that range, so treat them as the ceiling.

Under Cartesia pricing, what you pay per million characters depends on the plan and on whether you're inside the allowance.

Three speech extras draw from your credit balance too:

• Infill costs a fixed 300 credits per request, plus the standard rate on the infill transcript.

• Localizing a voice into another accent costs 225 credits per added accent.

• Each break tag in a transcript counts as 1 credit.

For a phone agent, latency counts alongside price. Cartesia lists sub-90ms model latency for Sonic-3.6 across 44 languages.

Our TTS for AI voice agents roundup puts Sonic next to seven other TTS engines for voice agents, with pricing for each.

Cartesia STT Pricing

Cartesia STT pricing depends on the model. You’ll pay 3 credits per second for Ink-2 over WebSocket, and for now it's streaming-only.

The older ink-whisper model handles both modes, at 1 credit per second streaming or 1 credit per 2 seconds in batch. Ink-2 currently covers five languages: English, Spanish, French, Hindi, and Japanese.

Source: Cartesia pricing docs list Ink-2 at 3 credits per second and ink-whisper streaming at 1 credit per second, and silence is billed even when no transcript comes back.

Cartesia's FAQ quotes $0.39 for an hour of Ink on Scale, while the rate card math comes to $0.40 (10,800 credits at $299 per 8M). On Pro, the same hour works out to $0.54.

Price per hour only matters if the transcripts hold up. On Cekura's STT benchmark, Cartesia Ink-2 scored a 2.39% word error rate on 1,000 Pipecat voice-agent clips, sixth of 17 models, and 3.09% on eight licensed Ocular call recordings, second of 17.

Its median final-text delay was 123ms, seventh of 17. The ranking already moves from sixth to second between two datasets, so re-run word error rate on a sample of your own calls before committing.

Cartesia Voice Agent Pricing

Cartesia pricing for Managed Agents is $0.06 per minute on every plan, plus $0.014 per minute when the call runs on a Cartesia-provided number, for $0.074 a minute before LLM usage.

Agent minutes come out of the prepaid balance first. Once it runs out, calls keep going at the same per-minute rate, so the balance covers your first minutes each month.

How far that balance goes depends on the plan. Before phone charges, Cartesia puts the included minutes at 17 minutes on Free, 1 hour 23 minutes on Pro, 13 hours 37 minutes on Startup, and 83 hours 3 minutes on Scale.

Two items sit outside that per-minute rate. Cartesia lists LLM usage during calls as free "for a limited time" for agents created in the UI, and evaluations as free for a limited time too.

For any other agent, Cartesia's docs say the per-minute rate doesn't include LLM usage, and token prices depend on the model.

Cartesia hasn't published an end date for either, so I'd budget as if both will start billing at some point.

Illustrative example: 10,000 three-minute calls a month is 30,000 agent minutes. That's $1,800 in call time plus $420 in telephony, or $2,220 before LLM usage on either Startup or Scale, because each plan's prepaid balance equals its fee.

So for this workload, the bill is the same on both plans, and the deciding limit is concurrency: 20 simultaneous calls on Startup and 60 on Scale.

Cartesia Pricing Limits: Concurrency, Seats, and Overages

Workspace seats are unlimited at all five tiers. Each tier limits concurrency, with a separate pool for TTS, STT, and agent calls. Per Cartesia's FAQ, TTS load doesn't count against STT or agent concurrency.

Overages run on a toggle on the subscription page, and only paid plans can use them. Turning them on keeps requests running past the allowance at your tier's overage rate.

Leave them off, and new requests fail once your credits are gone, until the next renewal or an upgrade. Spend stays capped, but any agent calling Sonic or Ink through the API gets errors in the meantime.

Once you're on Scale, I'd load-test its 15 TTS requests with all 60 calls running. Whether 15 is enough for 60 live callers depends on how many agents speak at the same moment, which only shows up at peak call volume.

Which Cartesia Plan Should You Choose?

To match Cartesia pricing to your usage, convert your monthly TTS characters and STT seconds into one credit total, then check that tier's concurrency limit at your peak call volume.

Choose Free if you:

• Want to hear Sonic voices and test Ink accuracy on audio you've recorded

• Don't need a commercial license yet

Choose Pro if you:

• Launch a small commercial feature with under 100K characters a month

• Want Instant voice cloning for one or two custom voices

Choose Startup if you:

• Run a branded voice made with Pro voice cloning

• Use between 100K and about 6.8M credits a month, paying overage past 1.25M

• Handle up to 20 concurrent agent calls

Choose Scale if you:

• Use more than about 6.8M credits a month (roughly 630 hours of Ink-2), where Startup plus overage passes $299

• Run up to 60 concurrent calls or STT streams

Choose Enterprise if you:

• Need a signed BAA, Zero Data Retention, or regional processing

• Plan to deploy on-prem or in a private VPC

Is Cartesia Worth the Cost?

On a fixed monthly workload of 1M characters of speech and 100 hours of streaming transcription, Cartesia pricing comes to $86.35 on Startup, the third-cheapest of four at list prices, behind Deepgram at $76.20 and ElevenLabs at $79. Full workings in the comparison section below.

In the example below, adding 100 hours of Ink-2 transcription to 1M characters of speech doubles credit use, from 1M to 2.08M.

Illustrative example: A voice product generates 1M characters of speech and transcribes 100 hours of streaming audio each month. On Ink-2, that's 2.08M credits. Startup covers 1.25M, and the remaining 830K cost $37.35 in overage, for $86.35 a month.

Switch transcription to ink-whisper, and that month uses 1.36M credits, for $53.95. Scale's allowance covers either version, at $299.

Cartesia is worth it if you:

• Want TTS, STT, and managed agents under one account, with TTS and STT sharing a credit pool

• Care about STT accuracy, since Ink-2 placed second of 17 on the eight Ocular call recordings in Cekura's STT benchmark and sixth on the 1,000 Pipecat clips

• Want to add teammates without paying per seat

Skip Cartesia if you:

• Require a BAA without signing an Enterprise contract

• Prefer a flat monthly tier between $49 and $299 over overage billing

Cartesia Pricing Compared With ElevenLabs, OpenAI, and Deepgram

Cartesia, ElevenLabs, OpenAI, and Deepgram bill speech on different units, so I priced one fixed monthly workload on each. Cekura is in the table too, but it prices voice testing rather than speech, so its row isn't a like-for-like comparison.

Each starting price uses that vendor's own unit, which means per voice testing minute for Cekura, per month for Cartesia, and per 1K or 1M characters for the TTS APIs. The workload table below puts all four speech providers on one monthly total.

Cartesia pricing and the alternatives' list prices were verified against vendor pages on October 8, 2026.

The workload matches the example above: 1M characters of TTS plus 100 hours of streaming STT, priced at list rates.

Price is only half of the comparison. On the eight Ocular call recordings in Cekura's STT benchmark, Ink-2 scored a 3.09% word error rate (second of 17), against 3.53% for GPT Realtime Whisper (fourth) and 4.39% for Deepgram Nova-3 (eleventh). ElevenLabs' Scribe models aren't in that benchmark.

For a wider field, our ElevenLabs alternatives guide compares Cartesia with four other options.

How Cekura and Cartesia Work Together

Cekura is a Cartesia partner that works at the testing layer, above the speech models Cartesia provides. Cekura tests the agent that runs on Sonic or Ink with simulated callers who interrupt, speak with different accents, or call from noisy places, and those callers can speak in Cartesia voices.

Cartesia provides: The speech layer itself, from Sonic voices and Ink transcription to Managed Agents on Cartesia phone numbers.

Cekura provides: Testing for the agent built on top, with simulated callers before launch and scoring of production calls after they end, whichever TTS or STT provider sits underneath.

Use both when: You run Sonic or Ink inside a production agent and change voices or prompts often. A swap from Ink-2 to ink-whisper cuts STT credits by two-thirds.

Cekura's Transcription Accuracy metric shows whether the cheaper model still gets names and numbers right: it compares Ink's transcript of each simulated caller turn with the words Cekura's caller actually spoke.

Pre-production

• A/B testing runs the same evaluators against two versions of your agent, such as before and after a Sonic voice or model change, and compares the runs metric by metric.

• Automated simulations use caller personalities that change accent and speaking speed, so transcription errors can show up in testing before launch.

Infrastructure

• Infrastructure testing measures per-turn latency and interruption handling, with background noise mixed in.

Observability

• Production monitoring scores each call you send to Cekura after it ends, for sentiment, the stage where the caller dropped off, and latency on every agent turn.

Native integrations cover Retell, VAPI, ElevenLabs, LiveKit, Pipecat, Bland, and more. Our guide to Cartesia text-to-speech covers how to test Sonic itself.

Cekura charges $30 a month per additional seat on pay-as-you-go, and seats are included on the Startup plan.

Cekura supports SOC 2, HIPAA, and GDPR compliance, covering transcript redaction, role-based access, and audit trails.

Pay-as-you-go starts with 300 free credits, about 60 minutes of voice testing, then $0.25 per voice testing minute. Cekura’s Startup plan costs $500 a month and includes a signed BAA.

My Bottom Line on Cartesia Pricing

Cartesia pricing stays at $5 or less on Free and Pro, and the next step after Startup's $49 is Scale at $299.

My pick changes with the workload:

• A prototype starts on Free and moves to Pro once you launch to customers.

• An early production agent goes on Startup, with overages on and a budget alert.

• High-volume transcription should go on Scale, or on ink-whisper inside Startup if its word error rate on your calls stays close to Ink-2's.

• Healthcare teams handling PHI need Enterprise, since Cartesia lists BAAs only there.

Whichever tier you pick, check TTS concurrency before the credit total. On Scale, 15 TTS requests can cap an agent handling 60 calls well before 8M credits run out.

Book a demo and run your Cartesia-powered agent through simulated callers before a voice or model change reaches production.

Frequently Asked Questions

How Much Does Cartesia Cost?

Cartesia pricing is $0 on Free, $5 a month on Pro, $49 a month on Startup, and $299 a month on Scale. Enterprise pricing is custom. Each paid plan includes model credits plus a prepaid agent balance equal to its price.

Is There a Free Version of Cartesia?

Yes, Cartesia AI pricing includes a free plan with 20K credits a month, about 27 minutes of TTS or 1 hour 51 minutes of STT, plus $1 in agent credit. Commercial use needs Pro or higher.

How Does Cartesia Calculate Credits?

Cartesia calculates credits by character for speech and by second for transcription. Sonic TTS takes about 1 credit per character, Ink-2 STT takes 3 credits per second, and ink-whisper takes 1 credit per second when streaming.

What Is Cartesia TTS Pricing per Minute?

Cartesia TTS pricing works out to about 750 to 800 credits per minute of audio. That's roughly $0.028 to $0.040 per minute within the Pro allowance through Scale.

What Is Cartesia STT Pricing per Hour?

Cartesia STT pricing for Ink-2 is 10,800 credits per hour of audio, about $0.40 on Scale and $0.54 on Pro. The ink-whisper model uses 3,600 credits per hour of streaming.

How Do You Test a Cartesia Agent Before Switching Models?

Run the same test calls against the agent before and after the switch, then compare the results. Cekura's simulated callers can speak in Cartesia voices, its Transcription Accuracy metric checks whether Ink still gets names and numbers right, and A/B testing compares the two runs metric by metric.

Cekura's pay-as-you-go plan starts with 300 free credits, about 60 minutes of voice testing.