ElevenLabs vs Cartesia (2026): Text-to-Speech and Voice Agent Pricing Compared

ElevenLabs vs Cartesia, checked 11 November 2026: TTS cost per million characters, hosted voice agents at $0.08 vs $0.03 a minute, languages, latency claims, concurrency, n8n and Vapi setup, with worked cost examples.

ElevenLabs vs Cartesia (2026): Text-to-Speech and Voice Agent Pricing Compared

Short answer: Choose Cartesia if you want the lowest per-minute cost for real-time voice agents: Sonic text-to-speech works out to about $37–$50 per million characters, and Managed Agents cost $0.03/min (+$0.014 on a Cartesia number) plus the LLM. Choose ElevenLabs if you need more languages (90+), more voices and a broader toolkit (dubbing, music, a verified n8n node): low-latency TTS is $0.04 per 1,000 characters, and hosted ElevenAgents cost $0.08/min plus the LLM. At 10,000 agent minutes a month that is about $440 vs $800 before LLM costs, by our arithmetic.

Checked 11 November 2026 (Asia/Tehran) against both pricing pages (rendered in a browser), Cartesia’s pricing docs and model pages, and ElevenLabs’ model docs. We haven’t benchmarked voice quality or latency ourselves.

Disclosure: some links in this article are affiliate links. If you buy through them, Sophrai may earn a commission at no extra cost to you. It never changes our rankings. Read our affiliate disclosure.

ElevenLabs and Cartesia are the two speech-model companies most voice-agent builders shortlist. Both sell a text-to-speech (TTS) API, a speech-to-text (STT) model and, since this year, a hosted voice-agent platform. They price those products in different units, though: ElevenLabs in characters and call minutes, Cartesia in credits and “agent dollars”. Below we put both on the same scale, cost per million characters and per call minute, and work through four real volumes.

ElevenLabs vs Cartesia at a glance

ElevenLabsCartesia
Flagship real-time TTSEleven v4 Turbo, Flash v2.5Sonic 3.6
TTS API price$0.04 per 1K characters (v4 Turbo, v3 Conversational, Flash/Turbo); $0.08 (v4, v3, Multilingual v2)1 credit per character; plans equal about $37–$50 per 1M credits; overage $38–$65 per 1M
Latency claim (vendor)~100 ms median (v4 Turbo, Flash v2.5), excluding app and networkFirst audio in under 90 ms
TTS languages90+ (v4, v4 Turbo); 32 (Flash v2.5); 29 (Multilingual v2)44 (Sonic 3.6)
Real-time STTScribe v3 Live $0.22/hour; Scribe v3 batch $0.20/hourInk 2: 3 credits per second (about $0.39/hour on Scale, per Cartesia)
Hosted voice agentsElevenAgents $0.080/min + LLMManaged Agents $0.03/min + $0.014/min telephony on a Cartesia number + LLM
Entry paid planStarter $6/mo ($1 first month until 16 Nov)Pro $5/mo
RolloverUp to 2 months (balance up to 3x monthly quota), paid plansUp to 2x monthly credits
n8nVerified node (built by ElevenLabs)HTTP Request node (no n8n integration page)
Vapi voice providerYes (`11labs`, up to Eleven v4 Turbo)Yes (`cartesia`, Sonic 3.6)

Sources: ElevenLabs API pricing, ElevenLabs models, Cartesia pricing, Cartesia pricing docs, Cartesia Sonic. All prices USD, monthly billing, excluding tax.

Text-to-speech pricing, side by side

ElevenLabs (API plans)

ElevenLabs’ API pricing tab converts each plan’s fee into included characters at the model’s rate. Annual billing is 10x the monthly price (two months free), per its FAQ.

PlanMonthlyAnnual (per month)Included characters, v4 Turbo / FlashIncluded characters, v4 / v3 / Multilingual v2TTS concurrency (Multilingual v2 / Flash)
Free$0$020,00010,0002 / 4
Starter$6 ($1 first month)$5150,00075,0003 / 6
Creator$22 ($11 first month)$18.33550,000275,0005 / 10
Pro$99$82.502,475,0001,238,00010 / 20
Scale (3 seats)$299$249.177,475,0003,738,00015 / 30
Business (10 seats)$990$82524,750,00012,375,00015 / 30

ElevenLabs estimates about 1,000 characters per minute of audio, so its page shows roughly $0.04 per minute for the low-latency models and $0.08 for the expressive ones. Its docs say the concurrency limit isn’t a cap on simultaneous conversations: as a rule of thumb, 5 concurrent requests can serve about 100 simultaneous conversations, because audio is generated faster than it’s played back.

Cartesia (credits)

PlanMonthlyCredits / monthEffective $ per 1M credits (our arithmetic)Overage per 1M creditsSonic TTS minutes (Cartesia’s estimate)TTS concurrency
Free$020,000–Not available~272
Pro$5100,000$50.00$65~1333
Startup$491,250,000$39.20$45~1,6675
Scale$2998,000,000$37.38$38~10,66715
EnterpriseCustomCustom–Volume pricingCustomCustom

Sonic TTS costs about 1 credit per character, roughly 750–800 credits per minute of audio by Cartesia’s estimate. Credits are spent only on successful requests. Overages are optional on paid plans; with them off, requests fail once credits run out. Unused credits roll over up to 2x your plan’s monthly credits.

Comparing per minute is misleading here: ElevenLabs assumes ~1,000 characters per minute and Cartesia ~750–800. Per million characters, ElevenLabs’ low-latency models cost $40 and Cartesia $37–$50 on plan, so for the same script the two are close. ElevenLabs’ expressive models (v4, v3) cost twice as much.

Voice agents: ElevenAgents vs Cartesia Managed Agents

ElevenLabs ElevenAgentsCartesia Managed Agents
Platform rate$0.080 per call minute (all plans)$0.03 per call minute (pricing page shows it cut from $0.06)
TelephonyNot charged by ElevenLabs; your provider bills you+$0.014/min on a Cartesia-provided number
LLMExtra, per-minute estimate by model (e.g. Gemini 2.5 Flash $0.0029/min in the calculator)Billed separately at the model’s token prices (via the models API)
Included usage15 / 75 / 275 / 1,238 / 3,738 / 12,375 min (Free to Business)Prepaid agent dollars equal to the plan fee: $1 / $5 / $49 / $299
Concurrent calls4 / 6 / 10 / 20 / 30 / 408 / 12 / 20 / 60 (Free, Pro, Startup, Scale)
Phone numbersYour own telephony provider (its charges are excluded from ElevenLabs’ prices)1 / 3 / 5 / 10 Cartesia-provisioned numbers
Other chargesText messages $0.003; burst calls $0.16/minEvaluations free “for a limited time”

Sources: ElevenAgents pricing, Cartesia pricing, Cartesia pricing docs, Cartesia Managed LLMs. For how ElevenAgents compares with an orchestration platform, see our Vapi vs ElevenLabs Agents breakdown; for Cartesia’s launch details, see Cartesia Managed Agents.

What it costs at real volumes

Our arithmetic from the list prices above (monthly billing, USD, before tax). TTS scenarios use characters so both vendors are on the same scale.

ScenarioElevenLabsCartesia
A. 1M characters of TTS a month (an agent or app speaking ~1,000–1,300 min)Low-latency models: $40 (Creator $22 incl. 550K + 450K at $0.04/1K, assuming extra usage bills at the same rate). v4 quality: $80Startup $49 (1.25M credits included)
B. 10M characters a monthLow-latency: Scale $299 (7.475M) + 2.525M x $0.04/1K = ~$400Scale $299 (8M) + 2M x $38/1M = $375
C. 1,000 agent minutes a month, hostedCreator $22 (275 min) + 725 x $0.08 = $80; or Pro $99 (1,238 min). Plus LLM and your telephony1,000 x ($0.03 + $0.014) = $44 (Pro $5 with overages on). Plus LLM
D. 10,000 agent minutes a month, hostedScale $299 (3,738 min) + 6,262 x $0.08 = ~$801; or Business $990 (12,375 min). Plus LLM (~$29 on Gemini 2.5 Flash) and telephony10,000 x $0.044 = $440 (e.g. Scale $299 + $141 overage, 60 concurrent calls). Plus LLM

The pattern: for raw TTS, the two are within about 10% of each other at volume, and model choice inside ElevenLabs (Flash vs v4) matters more than the vendor. For hosted agents, Cartesia’s platform rate is well under half of ElevenLabs’, but you still pay for the LLM, and with ElevenLabs you also pay a telephony provider. If you’d rather orchestrate your own stack, Vapi lets you plug in either voice (see the n8n section below).

Quality, languages and voices

  • Languages: ElevenLabs v4 and v4 Turbo list 90+ languages; Flash v2.5 32; Multilingual v2 29 (models). Cartesia Sonic 3.6 lists 44, including newly added Odia and Urdu (Sonic 3.6).
  • Latency: Cartesia says first audio arrives in under 90 ms. ElevenLabs lists a ~100 ms median for v4 Turbo and Flash v2.5 and ~165 ms for v3 Conversational, excluding application and network latency. Different methods, so not directly comparable.
  • Voice cloning: ElevenLabs offers instant cloning from Starter and professional clones from Creator (1 on Creator and Pro, 3 on Scale, 10 on Business). Cartesia offers instant cloning from Pro and professional clones from Startup (2 slots) and Scale (4); training a professional clone is free.
  • Model stability: Cartesia publishes dated snapshots (for example `sonic-3.6-2026-08-27`) that never change, which is useful if you run evals before every model update.
  • Speech-to-text: Cartesia’s Ink 2 is a streaming model with built-in turn detection but lists only 5 languages. ElevenLabs’ Scribe v3 Live costs $0.22 per hour with a multilingual mode, and its batch Scribe v3 lists 90+ languages at $0.20 per hour.
  • Breadth: ElevenLabs also sells dubbing, music, sound effects and a voice changer from the same credit pool. Cartesia is focused on real-time speech and agents.
  • Compliance: Cartesia says it is SOC 2 Type II certified; DPAs and BAAs are on Enterprise. ElevenLabs offers BAAs for HIPAA customers on Enterprise.

Using them with n8n and Vapi

ElevenLabs has a verified n8n node built and maintained by ElevenLabs: text to speech, speech to text, speech to speech, voice listing and cloning. Cartesia had no n8n integration page when we checked (the URL returned 404), so you call `POST /tts/bytes` with n8n’s HTTP Request node and store the returned audio as a binary file. For both, n8n is best for batch jobs (voice notes, audio digests, call summaries), not for streaming a live call.

For live phone agents, either use the vendor’s hosted agent platform or an orchestrator. Vapi lists both as built-in voice providers: Cartesia (`cartesia`, Sonic 3.6 and dated snapshots, docs) and ElevenLabs (`11labs`, from Flash v2 to v4 Turbo, docs). You can switch voices per assistant without changing your tools or webhooks. Estimate n8n execution volume with our n8n cost calculator, check tool-calling workflows with the n8n workflow auditor, and compare vendor prices in our AI SaaS pricing comparison.

Which should you choose?

  • Real-time voice agents on a budget: Cartesia. Managed Agents at $0.03/min, 60 concurrent calls on Scale and its own US numbers.
  • Multilingual products, content and dubbing: ElevenLabs. 90+ languages on v4, a larger voice library and one credit pool across audio products.
  • Automations in n8n: ElevenLabs, because the verified node saves you writing HTTP calls; Cartesia works fine via HTTP Request.
  • You want to swap voices later: build on an orchestrator like Vapi and A/B test Sonic against Eleven v4 Turbo on real calls. See also Vapi vs Retell vs Bland and our AI tools directory.

What we couldn’t verify

  • ElevenLabs overage on API plans. The API tab shows the same $0.04 / $0.08 per 1K rate on every plan and “usage based billing for additional credits” from Creator; we assumed extra usage bills at that rate.
  • ElevenLabs’ two meters. The creative tab describes the same plans in credits (for example 121K credits on Creator), while the API tab shows characters (550K Flash characters on Creator). Check which meter applies to your account.
  • “As low as 5c/minute”. ElevenLabs’ Business card says low-latency TTS “as low as 5c/minute”, which doesn’t match the $0.04/min API rate shown elsewhere on the same site.
  • v3 Conversational latency. The pricing page says ~280 ms; the model docs say ~165 ms. We quote the docs.
  • Cartesia LLM token prices. They are returned by the models API for logged-in accounts, not published, so we couldn’t add them to scenario D.
  • Whether Cartesia telephony draws on prepaid agent dollars. We assumed it does in our totals.
  • Voice quality and latency. All latency and quality claims are vendor figures; we didn’t benchmark either service.

Frequently asked questions

Is Cartesia cheaper than ElevenLabs?

For text-to-speech at volume, slightly. Cartesia’s plans work out to about $37 to $50 per million characters (Scale $299 for 8 million credits, 1 credit per character), and overage is $38 to $65 per million. ElevenLabs lists $0.04 per 1,000 characters ($40 per million) for its low-latency models (v4 Turbo, v3 Conversational, Flash/Turbo) and $0.08 ($80 per million) for v4, v3 and Multilingual v2. For hosted voice agents the gap is bigger: Cartesia charges $0.03 a minute plus $0.014 on its own phone number, ElevenLabs $0.08 a minute; both bill the LLM on top.

How much does ElevenLabs cost per minute of speech?

ElevenLabs’ API pricing page shows about $0.04 per minute for v4 Turbo and Flash/Turbo and about $0.08 per minute for v4, v3 and Multilingual v2, based on its own estimate of roughly 1,000 characters per minute. Its hosted ElevenAgents platform charges $0.080 per call minute on every plan, with included minutes from 15 (Free) to 12,375 (Business), plus LLM usage.

How much does Cartesia cost per minute?

Cartesia meters Sonic text-to-speech at 1 credit per character and says a minute of audio is about 750 to 800 credits. On the Scale plan ($299 for 8 million credits) that is about $0.028 to $0.030 per minute, by our arithmetic. Cartesia Managed Agents cost $0.03 per call minute plus $0.014 per minute on a Cartesia phone number; LLM usage during calls is billed separately.

Which is faster, ElevenLabs or Cartesia?

Both claim sub-100-millisecond results. Cartesia says Sonic gives first audio in under 90 ms. ElevenLabs’ model docs list a median of about 100 ms for v4 Turbo and Flash v2.5, excluding application and network latency. These are vendor figures measured differently, and we didn’t benchmark either, so test with your own voice, text and region.

Which supports more languages?

ElevenLabs. Its v4 and v4 Turbo models list 90+ languages (Flash v2.5: 32; Multilingual v2: 29). Cartesia Sonic 3.6 supports 44 languages natively, and its Ink 2 speech-to-text model currently lists 5 (English, French, Hindi, Japanese, Spanish).

Can I use ElevenLabs or Cartesia in n8n and Vapi?

Yes. ElevenLabs has a verified n8n node built by ElevenLabs (text to speech, speech to text, speech to speech, voices and cloning). Cartesia had no n8n integration page when we checked, so you call its REST API with n8n’s HTTP Request node. In Vapi, both are built-in voice providers: Cartesia Sonic 3.6 and ElevenLabs models up to v4 Turbo.

Related reading

Sources

ElevenLabs pricing (rendered); ElevenLabs API pricing (rendered, subscription view); ElevenAgents pricing (rendered); ElevenLabs models and concurrency docs; Cartesia pricing (rendered, incl. FAQ); Cartesia pricing docs; Cartesia Sonic 3.6; Cartesia Ink 2; Cartesia Sonic page; Cartesia Managed Agents; Cartesia Managed LLMs; n8n ElevenLabs integration; n8n HTTP Request node; Vapi: Cartesia voice; Vapi: ElevenLabs voice. All checked 11 November 2026 (Asia/Tehran). Scenario totals are our own arithmetic from list prices.

hamid7777 Avatar

Written & tested by

The Sophrai brief

One calm email when something actually matters.

New tools, real cost breakdowns and the automation news worth your time. No hype, no spam.