Cartesia Pricing in 2026: Plans Explained
Quick note before we start: Cartesia doesn’t run a public affiliate program that I could find, so the links in this article are plain links — I don’t earn anything if you sign up, and that doesn’t change a word of what follows. Every price here was read off Cartesia’s own pricing page on November 12, 2026. If a number looks different when you check, trust the live page over me.
I’ll be honest with you: the first time I opened Cartesia’s pricing page, I closed it again. Not because the prices were bad — they’re some of the cheapest in voice AI — but because the page talks about two different kinds of money at once. Credits for the speech models. Prepaid dollars for voice agents. Overage rates that change by tier. A free plan with a catch hiding in the small print. It reads like a phone bill written by engineers.
So I sat down, read the whole page twice, cross-checked it against a few independent pricing guides, and did the cost-per-minute math myself. This is the article I wish I’d had open the first time. Every plan, what the credits actually buy, where the traps are, and which plan makes sense for what you’re building.
The short version
If you just want the answer and you’re in a hurry, here it is:
- Free — $0 a month. 20,000 credits, roughly 27 minutes of speech. Fine for kicking the tyres. You can’t use anything you make commercially, so don’t build a product on it.
- Pro — $5 a month. 100,000 credits, roughly 133 minutes of speech, plus a commercial licence and instant voice cloning. This is the plan most people should buy, and it’s the cheapest serious voice plan I’ve found anywhere in 2026.
- Startup — $49 a month. 1.25 million credits, roughly 1,667 minutes, and Pro voice cloning with two slots. For a small team shipping a voice feature.
- Scale — $299 a month. 8 million credits, roughly 10,667 minutes, four Pro clone slots. For real volume — call centres, apps, agencies.
- Enterprise — custom. On-premise or private-cloud deployment, custom limits, talk to sales.
- The catch to know about: the old annual prices — $4, $39 and $239 a month — are gone. The page has no billing-period toggle anymore. $5, $49 and $299 is the price now, and that’s about 25% more than the cheapest rate you could buy a few months ago.
Want the detail? Keep reading. The rest of this article is the math behind those bullets.
Cartesia pricing at a glance
Cartesia sells three products under one roof: Sonic text-to-speech, Ink speech-to-text, and Managed Agents — hosted voice agents that take phone calls. One subscription covers all three, and each plan bundles two allowances: credits for the speech models, and prepaid dollars for the agents.
| Plan | Price | Credits / month | Sonic TTS minutes | Ink STT hours | Agent dollars |
|---|---|---|---|---|---|
| Free | $0 | 20,000 | ~27 | ~1h 51m | $1 |
| Pro | $5 | 100,000 | ~133 | ~9h 16m | $5 |
| Startup | $49 | 1,250,000 | ~1,667 | ~115h 44m | $49 |
| Scale | $299 | 8,000,000 | ~10,667 | ~740h 44m | $299 |
| Enterprise | Custom | Custom | Custom | Custom | Custom |
One thing I like about this setup: every paid plan hands back a prepaid agent balance equal to its own price. Pay $5, get $5 of agent calls. Pay $299, get $299 of agent calls. It’s a small thing, but it makes the plans easy to reason about — you’re never wondering where the agent money went.
How Cartesia credits actually work
This is the part that trips people up, so let’s get it straight before we talk about individual plans.
A credit is Cartesia’s unit of speech. For Sonic text-to-speech, one credit buys one character. A minute of spoken audio takes roughly 750 to 800 credits, because that’s how many characters a natural-sounding voice gets through in sixty seconds. So a 1,000-word blog post — call it 5,000 characters — costs about 5,000 credits. On the Pro plan, that’s twenty blog posts a month, if you used nothing else.
Ink speech-to-text is priced differently: 3 credits per second of audio, which works out to $0.39 per hour on the Scale plan. Transcription is cheap here. Nine hours of it comes bundled with Pro, and more than a hundred hours with Startup.
A few details from the pricing FAQ that matter in practice:
- Break tags cost 1 credit each. If you pepper your scripts with pauses, they add up — slightly, but they add up.
- Credits roll over, up to twice your monthly allowance. On Pro that means you can hold up to 200,000 credits. Unused credits don’t vanish at the end of the month, which is more generous than most voice tools.
- Changing tiers keeps your credits. Upgrade and the new tier’s credits land in your balance straight away. Downgrade or cancel mid-cycle and you keep what you paid for until you use it — your account just drops to the new tier at the end of the billing period, and the rollover cap resets to twice the new plan’s allowance.
- Overages are off by default. On paid plans, once your credits hit zero, new requests simply stop — unless you’ve switched overages on. That’s the safe default, and I wish more companies did it this way.
- Voice clones speak at the standard rate. Training a Professional Voice Clone is free, and speech from a clone costs 1 credit per character, the same as any other voice.
Here’s the mental model I ended up with: credits are your fuel, the plan is the size of your tank, and the rollover cap is how much fuel you’re allowed to keep in the garage. Once that clicks, the rest of the page makes sense.
Free plan: $0 a month
The free plan gives you 20,000 credits a month — about 27 minutes of speech or just under two hours of transcription — plus $1 of voice-agent credit, which covers roughly 33 minutes of calls. You get two concurrent text-to-speech requests and eight concurrent transcription requests, and one Cartesia-provisioned phone number.
That sounds generous, and for prototyping it genuinely is. Twenty-seven minutes is enough to wire up the API, test a voice, build a demo and show it to someone. No card required, which is how a free tier should work.
But there’s a catch, and it’s a big one: the free plan has no commercial licence. Anything you generate on it can’t be shipped in a product, sold, or used in monetised content. It’s a sandbox, not a launchpad. I’ve seen people build whole demos on free tiers and only discover the licence problem when they’re about to go live — don’t be that person.
Two more things to know. First, the free plan has no overages at all — when the credits are gone, you’re done until next month. Second, instant voice cloning used to sit on the free tier earlier in 2026; on the current pricing page it’s listed from the Pro plan up. If you’re reading an older guide that tells you cloning is free, that guide is out of date.
Who it’s for: developers testing the API, anyone comparing voices before committing, and students. If you’re making anything for real users, move on to Pro — it’s five dollars.
Pro plan: $5 a month
Pro is the plan this whole article is really about, because it’s where Cartesia stops being a demo and starts being a tool.
For $5 a month you get 100,000 credits — about 133 minutes of Sonic speech or 9 hours and 16 minutes of transcription — plus $5 of prepaid agent credit, worth roughly 2 hours and 47 minutes of voice-agent calls at the standard $0.03-a-minute rate. Concurrency goes up to 3 simultaneous text-to-speech requests, 12 transcription requests and 12 concurrent calls, with 3 phone numbers included.
And this is the tier where the two things that matter most unlock: the commercial use licence and instant voice cloning. Upload a short audio sample — Cartesia says about 10 seconds is enough — and you can generate speech in that voice across 42 languages. At five dollars, that’s the cheapest voice cloning on the market by a wide margin. ElevenLabs’ entry plan costs a dollar more and gives you less than a third of the characters.
Let’s do the cost math, because this is where Pro shines:
- About $0.04 per minute of speech. Five dollars divided by 133 minutes is $0.0376. Call it four cents.
- $50 per million characters. That’s the effective rate if you use the full allowance.
- One blog post’s worth of narration costs about $0.25. Five thousand characters at $50 per million. A ten-minute YouTube voiceover runs you about forty cents.
Pro has one weakness: the overage rate. If you switch overages on and blow past your 100,000 credits, extra credits cost $65 per million — which is 30% more than the $50-per-million rate baked into the plan. So overages on Pro are a bad deal by design. My advice: leave them off, let requests fail at zero, and upgrade to Startup if you’re regularly running dry. We’ll come back to that in the overages section.
Who it’s for: YouTubers and podcasters narrating their own scripts, indie hackers adding a voice to a side project, small creators who need a commercial licence and a clone of their own voice. If you’re generating under two hours of speech a month, this is your plan — and honestly, it’s hard to argue with five dollars.
Startup plan: $49 a month
Startup is a 12.5x jump in credits for roughly 10x the price, which tells you something about how Cartesia thinks about scale: the per-unit price gets better as you go up, but not dramatically. You get 1.25 million credits — about 1,667 minutes of speech or nearly 116 hours of transcription — plus $49 of agent credit, good for around 27 hours of calls.
What actually changes at this tier isn’t just volume. Startup is where Pro voice cloning unlocks — the higher-fidelity clone, with two slots included. Instant cloning copies a voice’s characteristics; Pro cloning trains a proper model on more audio and sounds noticeably closer to the real thing. If your product’s whole pitch is “it sounds like a real person,” this is the tier that delivers it. Training is free; speech from a Pro clone still costs the standard 1 credit per character.
Concurrency jumps too: 5 simultaneous text-to-speech requests, 20 transcription requests, 20 concurrent calls, 5 phone numbers. That’s a small team running a live product, not one person narrating videos.
The cost math:
- About $0.03 per minute of speech. $49 divided by 1,667 minutes is $0.0294.
- $39.20 per million characters — about 22% cheaper per character than Pro.
- Overages drop to $45 per million, which is much closer to the plan rate than Pro’s $65. Overage billing starts to make sense here.
There’s also a subtle perk buried in the feature list: localising a voice into another accent costs 225 credits per added accent, and Startup is the tier where managing multiple voices and accents starts to feel practical rather than fiddly.
Who it’s for: startups shipping a voice feature — a support agent, a reading app, a language tool — and creators whose monthly output has outgrown Pro. If you’re burning through 100,000 credits before the month is half over, the upgrade pays for itself in overage savings alone.
Scale plan: $299 a month
Scale is Cartesia’s top self-serve tier, and the numbers get serious: 8 million credits a month, about 10,667 minutes of speech or nearly 741 hours of transcription, $299 of agent credit covering roughly 166 hours of calls, 15 concurrent text-to-speech requests, 60 transcription requests, 60 concurrent calls and 10 phone numbers.
Pro voice cloning goes up to four slots. Overage credits fall to $38 per million — almost exactly the plan’s own effective rate of $37.38 per million characters, so overages stop being a penalty and start being a fair extension of the plan. At this level, the pricing is basically linear: you pay for what you use, and the plan is just a volume commitment.
The per-minute math flattens out too. At $299 for 10,667 minutes, you’re paying about $0.028 a minute — barely cheaper than Startup’s $0.029. The jump from Startup to Scale isn’t about a better rate; it’s about headroom. Fifteen concurrent TTS requests and sixty concurrent calls is call-centre territory.
One detail worth noting: the pricing FAQ lists transcription at $0.39 an hour on Scale, which is the cheapest Ink rate on the page. If you’re transcribing hundreds of hours — meeting notes, call recordings, podcasts — Scale’s transcription alone can justify the price against a dedicated transcription tool.
Who it’s for: companies running voice agents at real volume, agencies reselling voice capacity, and apps where speech is the product rather than a feature. If you’re reading this section wondering whether you need it, you probably don’t — yet.
Enterprise: custom pricing
Above Scale there’s Enterprise, priced by sales. The page doesn’t publish numbers, which is normal for this tier. What it does promise: custom credit and concurrency limits, on-premise or virtual-private-cloud deployment of the models, and custom terms around Pro voice cloning slots.
The on-premise option is the interesting one. Cartesia confirms you can deploy the models inside your own infrastructure, which matters for healthcare, finance and government work where audio can’t leave your network. If that’s your situation, you’re already talking to sales — this article isn’t going to change your mind either way.
Who it’s for: regulated industries, very high-volume deployments, and anyone who needs the models running inside their own cloud.
Managed Agents: the second bill hiding inside your plan
Here’s the part most pricing explainers skip, and it’s the part that can surprise you. Every plan includes a prepaid agent balance — $1 on Free, $5 on Pro, $49 on Startup, $299 on Scale — and Managed Agents burn through it at $0.03 a minute for call duration, plus $0.014 a minute in telephony charges when the call runs on a Cartesia-provided phone number. LLM usage during calls is billed separately on top.
Cartesia’s own FAQ translates the prepaid balances into minutes, and the numbers are worth keeping:
| Plan | Agent dollars | Call minutes included |
|---|---|---|
| Free | $1 | ~33 minutes |
| Pro | $5 | ~2 hours 47 minutes |
| Startup | $49 | ~27 hours 13 minutes |
| Scale | $299 | ~166 hours 7 minutes |
Those minutes don’t include telephony or the language model, so a real call costs a bit more than $0.03 a minute all-in. If you’re running an agent on your own phone number, you skip the $0.014 telephony charge; if Cartesia provisions the number, you pay it. Evaluations are free of charge for a limited time, according to the page — I’d read “limited time” as “don’t build your budget around this staying free.”
The practical point: if you’re only doing text-to-speech, the agent balance is irrelevant — it just sits there. If you’re building a voice agent, the agent balance is half your bill, and you should size your plan by call minutes, not by credits. A Pro plan gives you 133 minutes of TTS but only about 167 minutes of calls. Those are different budgets, and confusing them is the easiest way to get an unexpected invoice.
Overages work the same way as credits: off by default on paid plans, and once they’re on, usage past your balance is charged at the same rates. You need a payment method on file, and overages are billed when your plan renews.
The overage trap (and how to dodge it)
Overage pricing is where Cartesia’s plans quietly punish small users and reward big ones. Here’s the comparison:
| Plan | Effective rate (per 1M credits) | Overage rate (per 1M credits) | Overage premium |
|---|---|---|---|
| Pro | $50 | $65 | +30% |
| Startup | $39.20 | $45 | +15% |
| Scale | $37.38 | $38 | +2% |
On Pro, every overage credit costs 30% more than a credit inside the plan. Run 74,000 credits over in a month and you’ve paid $4.81 extra on top of your $5 — nearly doubling the bill for 74% more usage. That’s the trap: Pro is cheap if you stay inside it and expensive the moment you don’t.
Three ways to dodge it. First, leave overages off and let requests fail at zero — the default, and the safest. Second, use the rollover: credits stack up to twice your monthly allowance, so a quiet month banks capacity for a busy one. Third, upgrade before you overage. If you’re regularly burning past 100,000 credits, Startup’s $49 buys 1.25 million of them at a better rate than Pro’s overages ever will.
On Scale, honestly, don’t worry about it. A 2% premium is a rounding error.
What happened to the annual discount?
If you read older Cartesia pricing guides, you’ll see different numbers: Pro at $4 a month, Startup at $39, Scale at $239. Those were real — the pricing page used to show lower per-month figures alongside the $5, $49 and $299 prices, with the cheaper numbers working out about 20% lower.
They’re gone now. The current pricing page has no billing-period toggle at all, and the app quotes the same $5, $49 and $299 as the page. So the lowest price you can buy Cartesia for is 25% higher than the lowest price you could buy it for a few months ago. That’s not a small change, and it’s worth saying plainly because older articles — including some of mine — still show the old numbers.
Does it change the verdict? Not really. Five dollars is still the cheapest entry point for commercial voice AI with cloning that I’ve found, and the per-minute rates are still a fraction of ElevenLabs’. But if you were waiting for an annual plan to lock in the old prices, that door has closed. The price is the price now, month to month, pause or cancel anytime.
Cartesia vs ElevenLabs on price
Most people reading this are choosing between Cartesia and ElevenLabs, so let’s put the numbers side by side. I covered ElevenLabs’ pricing in detail in my ElevenLabs pricing breakdown, and the full head-to-head is in my Cartesia vs ElevenLabs comparison — here’s the short version.
| Cartesia Pro | ElevenLabs Starter | ElevenLabs Creator | |
|---|---|---|---|
| Price | $5/mo | $6/mo ($5 annual) | $22/mo ($18.33 annual) |
| Credits | 100,000 | 30,000 | 121,000 |
| Cost per minute | ~$0.04 | ~$0.20 | ~$0.18 |
| Cost per 1M chars | $50 | $200 | ~$182 |
| Voice cloning | Instant, from Pro | From Starter | From Starter |
| Commercial licence | Yes | Yes | Yes |
Cartesia Pro gives you more than three times the characters of ElevenLabs Starter for a dollar less. That’s not a close call — it’s a rout on price. ElevenLabs earns its premium elsewhere: a bigger voice library, more mature dubbing and studio tools, and a community that shares voices. If you want the best-sounding voice for a YouTube narration and you don’t mind paying for it, ElevenLabs still has a case. If you want the cheapest way to generate a lot of good speech, or you’re building a voice agent where latency matters, Cartesia wins on the numbers and it isn’t close.
One more comparison for context: Murf’s Creator plan is $29 a month ($19 annual) for a similar seat at the table. I broke Murf’s pricing down in my Murf pricing explainer. Cartesia undercuts both of them, which is exactly why it’s become the default recommendation for developers in 2026.
Which Cartesia plan should you pick?
Forget the feature tables for a minute. Pick by what you’re actually doing:
- Testing the API or comparing voices? Free. Twenty-seven minutes is plenty, and you haven’t spent anything.
- Narrating videos, podcasts or audiobooks? Pro. At four cents a minute, a ten-minute video costs forty cents to voice, and the commercial licence covers you.
- Building a voice agent or a support bot? Pro to start, Startup when your call minutes pass ~3 hours a month. Size by agent dollars, not credits — that’s the budget you’ll actually hit first.
- Transcribing hours of audio? Pro handles about nine hours a month. Past that, Startup’s 116 hours is the sensible step.
- Shipping a product where the voice is the feature? Startup, for Pro voice cloning. Instant cloning is good; the trained Pro clone is noticeably better, and two slots cover a main voice plus a backup.
- Running a call centre or an agency? Scale. Sixty concurrent calls and 8 million credits is infrastructure, not a subscription.
- Unsure between two plans? Start lower. Credits roll over up to twice your allowance, so a quiet month on Pro banks capacity — and upgrading mid-cycle adds the new credits immediately. There’s no penalty for starting small.
And if you’re looking for a genuinely free voice for commercial use, not just a trial — that’s a different question, and I wrote a whole guide to it: 7 free text-to-speech tools with commercial use. Cartesia’s free tier isn’t on that list, because of the licence. Worth knowing before you build on it.
Frequently asked questions
Is Cartesia really only $5 a month?
Yes. The Pro plan is $5 a month, billed monthly, with 100,000 credits included. There used to be a cheaper annual rate of $4 a month, but Cartesia removed the billing-period toggle, so $5 is the price now. You can pause or cancel anytime.
How many minutes of speech does 100,000 credits buy?
About 133 minutes, according to Cartesia’s own estimate. That’s because one credit buys one character of speech, and a minute of natural audio runs about 750 to 800 characters. A 1,000-word script is roughly 5,000 characters, so Pro covers about twenty of those a month.
Do Cartesia credits roll over?
Yes, up to twice your monthly allowance — 200,000 credits on Pro, 2.5 million on Startup. You keep credits you’ve paid for even if you downgrade or cancel; the cap just resets to twice the new plan’s allowance. If you upgrade mid-cycle, the new tier’s credits are added to your balance right away.
Can I use the free plan commercially?
No. The free plan has no commercial licence, so anything you generate on it is for prototyping and testing only. The commercial licence starts on the $5 Pro plan. If you need a free commercial voice, see my guide to free TTS tools with commercial use.
Is there a Black Friday deal for Cartesia in 2026?
Not that I could find. I checked Cartesia’s pricing page and announcements on November 12, 2026, and there’s no Black Friday offer listed — no discount code, no sale banner, nothing. Given they just removed the annual discount, I wouldn’t hold my breath for a seasonal one either. Five dollars is already the cheapest commercial voice plan I’ve found, so there’s not much room for a sale to cut into.
Does Cartesia have an affiliate program?
Not a public one I could find. There’s no affiliate page on their site and no program listed in the usual directories. That’s why the links in this article are plain links. If that changes, I’ll update this section.
The verdict
Cartesia’s pricing is the rare kind that’s simple once you understand the two currencies. Credits for speech, dollars for agents, rollover up to twice your allowance, overages off by default. The plans are honest: $0 to test, $5 to ship, $49 to scale a product, $299 to run one at volume.
The $5 Pro plan is the story here. A hundred thousand credits, a commercial licence and instant voice cloning for less than a coffee — at about four cents a minute, it’s the cheapest serious voice AI you can buy in 2026, and the gap to ElevenLabs is wide enough that price alone settles it for most people. The trade-off is real — smaller voice library, fewer studio tools, no annual discount anymore — but if your question is “how much will this cost me,” the answer is refreshingly small.
My advice: start on Free for an afternoon, see if the voices fit your project, then pay the five dollars. And if you’re building a voice agent rather than narrating scripts, watch the agent dollars, not the credits — that’s the budget that will sneak up on you.
Prices checked November 12, 2026, against cartesia.ai/pricing. Cartesia has already changed its prices once this year, so re-check at checkout before you pay.
References
- Cartesia pricing page — cartesia.ai/pricing (checked November 12, 2026; plan prices, credit allowances, concurrency, agent rates, overage rates, rollover and FAQ all read directly from this page)
- Cekura — “Cartesia Pricing 2026: Every Plan, Credit & Overage” (independent guide, verified against Cartesia’s pricing page October 8, 2026)
- AI Alleyway — “Cartesia review 2026: the fastest voice AI, free to $299” (confirms the $4/$39/$239 annual figures are gone and both surfaces now read $5/$49/$299)
- Real AI Test — “ElevenLabs vs Cartesia (2026): The Price, Latency, and Clone Tradeoffs” (ElevenLabs comparison figures, checked September 26, 2026)
- Affitrove — Cartesia vs ElevenLabs in 2026; ElevenLabs Pricing in 2026: Credits Explained; Murf Pricing in 2026: Plans Explained
