Cartesia

Published September 3, 2026 · This page indexes everything this site has verified about Cartesia and links to where each finding is set out in full. It is not a review — the review is. No listening test has been run, so nothing here says how Cartesia sounds.

On this page

What it costs

Plans as displayed on cartesia.ai/pricing, verified September 1, 2026: Free $0/month with 20K credits (~27 minutes, no commercial licence, no cloning); Pro $5/month with 100K credits (~133 minutes, adds a commercial licence and instant cloning); Startup $49/month with 1.25M credits (~1,667 minutes, plus 2 professional clones); Scale $299/month with 8M credits (~10,667 minutes, plus 4 professional clones); Enterprise custom. Billing is monthly only — the page carries no annual toggle and no yearly price. The minute figures are Cartesia’s own approximations for Sonic 3.6, and the site’s computed per-minute rates from them are roughly $0.038 at Pro, $0.029 at Startup and $0.028 at Scale, a rate that barely improves across a sixty-fold jump in volume (the review). pricing comparison (read September 1, 2026) normalizes the published overage rates of $65, $45 and $38 per million credits to $65.00, $45.00 and $38.00 per million characters using Cartesia’s own factor — "Standard TTS costs approximately 1 credit per character" — and computes the plan divisions at full allowance use as $50.00 (Pro), $39.20 (Startup) and $37.38 (Scale) per million characters; it also records that the answers in Cartesia’s pricing-page FAQ, where that conversion lives, are not in the rendered document. Agent usage is a second meter entirely — "Cartesia meters model usage in credits and agent usage in agent dollars" — billed per minute in USD at $0.06 for a base call. Commercial use is prohibited by default under Terms §4.1 and the only official statement tying it to a price is the "Commercial use license" bullet on the $5 Pro card; no Cartesia page defines that licence’s scope or whether it survives cancellation.

Every rate above is set against every other provider’s, normalized to one unit, in our pricing comparison.

Where Cartesia’s own material disagrees with itself

These are contradictions inside the vendor’s published documentation, not disagreements between us and the vendor. We report both readings rather than choosing one, and each links to the page where we set it out in full. We have published 11 in total; the 10 with the most bearing on a decision are below, and the rest sit in the pages linked underneath.

• Cartesia publishes a marketing page at cartesia.ai/product/ai-dubbing headlined "Fastest multilingual AI video dubbing", whose numbered how-to instructs you verbatim to "Upload your video content and select the desired languages for dubbing using Cartesia’s intuitive platform." No such upload exists in any Cartesia documentation: there is no dubbing endpoint, no video ingestion, no dubbing parameter and no dubbing entry in its pricing. Cartesia’s documented products are text-to-speech, speech-to-text and voice agents. Where we published this.
• The same dubbing marketing page claims a "time-to-first-audio of just 40ms" against Cartesia’s own documentation, which describes "about 90ms" for Sonic 3.5. Where we published this.
• The same dubbing marketing page advertises cloning "from a 3 second clip" against Cartesia’s own documentation, which states "as little as 10 seconds of audio". Where we published this.
• The Voice Changer endpoints POST /voice-changer/bytes and POST /voice-changer/sse are listed in Cartesia’s breaking-changes table with a sunset date of August 20, 2026 and an em-dash rather than a replacement — twelve days before the site’s check — while Cartesia’s marketing site still publishes a Voice Changer capability page and links it from the footer. Cartesia does not say which is current. Where we published this.
• Terms of Service clause 4.1 permits "personal, non-commercial use only (unless commercial use is expressly permitted by your subscription tier)", and the contract never says which tier qualifies. The only official statement connecting that permission to a price is a feature bullet — "Commercial use license" — on the $5 Pro card of the pricing page, which is not a licence grant. Where we published this.
• Cartesia advertises "44 languages". The site counted 44 codes in the documentation but 46 names in the marketing list — and the marketing list not only holds two more, it drops Croatian and Gujarati (read September 2, 2026). Separately, Cartesia’s own locale table covers just 5 of its 44 languages. Where we published this.
• Cartesia’s headline "sub-90ms latency" names no metric on any Cartesia page, while a separate documentation landing page describes Sonic 3.5 as streaming "the first byte of audio in about 90ms" — two different quantities sharing one number. Meanwhile Sonic 3.6, the model Cartesia announced as generally available, carries no latency figure at all: its page contains zero values of the form "nms". The headline number belongs to the previous model. Where we published this.
• Two Cartesia pages attribute its text-normalization behaviour to different model generations, and the site records whether that behaviour belongs to Sonic 3.5 or Sonic 3.6 as unresolved. Where we published this.
• There is a gap between Cartesia’s policies and its product on consent. The Acceptable Use Policy clause 2 requires that you "may only submit your own voice and audio recordings or those of others with explicit consent", and Terms clause 4.2 separately prohibit making available a voice without that person’s express permission — yet the Clone Voice API reference contains no occurrence of the word "consent" and has no attestation field. The request takes a clip, a name, a language and descriptive metadata; the obligation is a warranty you give, not a gate the product enforces. Where we published this.
• Cartesia’s documentation says its agent "handles turn-taking and interruptions" in prose, but interruption is not exposed as a configurable field in the agent configuration reference, and its rich turn-detection parameters are parameterised on the standalone speech-to-text endpoint rather than on the agent product. Where we published this.

Dated and time-sensitive

Facts with a clock on them — model sunsets, policy revisions, market status. These are the entries most likely to be out of date first, which is why they carry their dates.

• 2026-10-20. sonic-2 and sonic-turbo stop working after this date at alias level — all snapshots, so pinning a dated version does not protect you — along with the sonic-3-2025-10-27 snapshot only. Documented migration target is Sonic 3.6. Corroborated on three official pages (the API-changes "Upcoming sunsets" table, the older-models page and the July 2026 changelog entry), all checked September 1, 2026. Requests to a retired model return an error carrying "error_code": "model_sunsetted". Source on this site.
• 2026-08-20. Published sunset date for the Voice Changer endpoints POST /voice-changer/bytes and POST /voice-changer/sse, with no replacement listed. The date had already passed twelve days before the site’s September 1, 2026 check, while the marketing capability page remained live; the site records whether the endpoints are actually off as unverified. Source on this site.
• 2026-06-01. Executed sunsets: the earlier model generation — sonic, sonic-english, sonic-multilingual and two dated snapshots — was retired in favour of newer Sonic, and the voice-embedding endpoints /voices/clone/clip, /voices/mix and POST /voices were retired in favour of voice IDs. Source on this site.
• 2024-06-14. Last-revised date displayed on both Cartesia’s Terms of Service and its privacy policy — over two years old at the time of check, and containing no model names and no lifecycle language. Source on this site.
• 2025-07-23. Last-revised date of Cartesia’s Acceptable Use Policy — the only one of its three governing documents revised since June 2024. It carries the clause 2 consent warranty. Source on this site.
• 2026-08. Cartesia’s August 2026 changelog states that migration is manual: "For customers currently using sonic-3.5, you will have to manually repoint production traffic to sonic-3.6-2026-08-27 or sonic-3.6." The same entry reassures that "Instant and Professional Voice Clones work on Sonic 3.6 with the same voice IDs — nothing to recreate" and that you "Keep your existing voice IDs, request shape, and prompting — no code changes required." Source on this site.
• 2026-09-02. This site published a correction concerning Cartesia: its flagship comparison had stated in two places that Cartesia claims "sub-90 ms to first audio", citing the pricing page. Re-checked on September 2, 2026, that pricing page contains no latency figure at all — no occurrence of "latency" and no millisecond value — and where Cartesia does publish "sub-90ms latency" it names no metric. Both occurrences were corrected. Source on this site.
• 2026-09-01. First publication of the Cartesia review, with the deprecation table and pricing page re-fetched and re-confirmed the same day. No listening test, no measured latency and no numeric score appear on the page; Cartesia’s marketed real-time performance is not restated as a finding. Source on this site.

Every page on this site about Cartesia

• Full review. Cartesia Review 2026: Pricing, Sunsets and Licensing, Docs-Verified — the complete documentation-based assessment, with sources and dates against every claim.
• Alternatives. Cartesia alternatives — organized by the reason you would leave, not as a ranked list.
• Head to head. Cartesia vs MiniMax, Cartesia vs OpenAI.
• In our buying guides. Best AI Voice Generators in 2026: 10 Providers Compared, Best AI Dubbing Software 2026: Only Four Vendors Actually Sell It, Best Free AI Voice Generators 2026: What “Free” Actually Buys You, Most Realistic AI Voices: What Every Vendor Charges for the Claim, Best Multilingual AI Voice Generators: Every Count, Recounted, Real-Time AI Voice: Why the Latency Numbers Cannot Be Compared, Best AI Voice Cloning Software 2026: Price, Access and Consent Compared.
• In our comparisons of published data. AI Voice Generator Pricing Comparison: Every Rate Normalized, Best Text-to-Speech APIs: Limits, Transports and Auth Compared.
• In our explainers. Can You Use AI Voices Commercially? What Ten Vendors Actually Say, How AI Voice Cloning Works: What Ten Vendors Require of You, How AI Voice Generators Work, According to the Vendors Themselves.
• By what you are making. AI Voice for Audiobooks: Will Chapter 20 Match Chapter 1?, AI Voice for Business: What Survives a Procurement Review, AI Voice for eLearning: Courses Change, and the Audio Must Follow, AI Voice for Podcasts: Only Three Providers Do Two Speakers, AI Voice for YouTube: What You Must Disclose Before You Publish.

What we could not verify

Honesty about gaps beats a page that looks complete. Standing limits on everything above, as of September 3, 2026:

• How Cartesia sounds. Our listening benchmark is designed but not running. Nothing on this page or on any page it links to ranks Cartesia on audio quality, naturalness or speed.
• Anything Cartesia does not publish. Every page indexed here is built from the vendor’s own documentation. Where that documentation is silent, we record the silence rather than estimating around it.
• Whether a documented feature works as documented. We read documentation; we did not create an account or make an API call.
• Currency. Each linked page carries its own check date. This index inherits them and does not re-verify on load.

Change log

September 3, 2026 — First publication. Assembled from pages already published on this site; no new vendor claim was introduced by this page.

Published September 3, 2026 · This page shows no audio and reports no listening results, because our benchmark has not run. Corrections are recorded with their date on the corrections page. Independence note: this page contains no affiliate links, and no vendor paid for placement or influenced the order — see how we make money and our editorial policy.