Skip to content
All master reviews

Review Machine · Master Review · AI platforms

ElevenLabs

One account, the whole audio pipeline

Across six ElevenLabs pages, two blind-listener boards, seven review sites and blogs, two App Store headline ratings and one public Hacker News figure read on 19 September 2026, the platform's breadth is what the reviews mention in passing and its voice quality is what they lead with, while four of the seven report paying for regenerations they did not use.

Plans listed Free $0 to Business $990Credits a month 10,000 to 6,000,000Models listed Eleven v3, Multilingual v2, Flash v2.5Read 19 September 2026Skip to the verdictRead as text

Verdict age

ProvisionalReview sites in, owners uncounted

The platform moves monthly. The boards date Flash v2.5 to December 2024, Multilingual v2 to August 2023 and Eleven v3 to February 2026, and its own pages carried a command-line client in August 2026. The review sites are dated 29 April to 1 August 2026 or undated, and the prices are read on 19 September 2026, so any verdict here is short-lived.

  • Early · editors only
  • Provisional · editors in, owners thin
  • Settled · owner reviews read over months

From 18 sources to one verdict

Seven levels, in the same order in every Master Review. Each line is the takeaway; open a level for the evidence under it, or stop when you have enough.

What we read

18 sources, 3 Sep 2025 – 19 Sep 2026. Editorial, buyer, forum and other tiers thin.

By tier, heaviest first

T1Official6

Pricing, models, voice cloning, safety, the Prohibited Use Policy and the Terms of Use, read 19 September 2026.

T2Editorial2Thin

Two Artificial Analysis blind-listener boards, read 19 September 2026. No established publication's scored review of the platform was readable: that tier carries news about the company, not reviews of the product.

T2News0None readable
T3Buyers2Thin

Two App Store headline ratings, one request each, read 19 September 2026. No buyer review was read one by one.

Forum1Thin

One public Hacker News figure, 409 stories matching the name, read 19 September 2026. The one thread read discussed the company's name and market and carries no finding here.

Other7Thin

Seven review sites and creator blogs. Four are dated 29 April to 1 August 2026; three carry no date of their own and are dated the day read.

Date window

3 Sep 2025 to 19 Sep 2026

Editorial reviews 19 Sep 2026Official pages, boards and store figures read 19 September 2026; the review sites and blogs from 29 April to 1 August 2026 or undated.

Couldn’t read, so not used

  • TechRadar explainerempty page
  • Utilo reviewempty page behind its navigation

Neither page was counted and nothing here rests on either.

Read, then rejected

  • Speechify's alternatives page · a competing vendor's own marketing, not a review
  • CrowdVerdict aggregate · an AI-synthesised score over posts we could not read ourselves

What they measured

Eleven v3 sits fifth on the controlled-voice board at $100 per million characters; Flash v2.5 is $50.

Credits a month, entry plans

10,000creditsFree, $0

ElevenLabs · read 19 Sep

121,000creditsCreator, $22

ElevenLabs · read 19 Sep

600,000creditsPro, $99

ElevenLabs · read 19 Sep

API price per million characters

Artificial Analysis · read 19 Sep

Blind-listener Elo, controlled voices

1048Eleven v3, rank 5

Artificial Analysis · read 19 Sep

1117Cartesia Sonic 3.6, rank 1

Artificial Analysis · read 19 Sep

Headline ratings

4.84/520,527 ratings, voice generator appHeadline only

App Store · read 19 Sep

4.72/510,278 ratings, ElevenReaderHeadline only

App Store · read 19 Sep

First-take generation time

12.5s991 characters on eleven_multilingual_v2, 3 July 2026

ToolProven · updated 19 Jul

Where they agree

The voice is what the review sites lead with; the breadth of what the subscription covers is what they mention on the way out.

One tick per editorial review, in order of publication:AA Artificial AnalysisAA2 Artificial Analysis

One account covers the whole audio pipeline

0/2

Mentioned in a clause by six of the seven review sites and rarely in a headline. ElevenLabs' pricing FAQ lists Speech to Text at 330 credits a minute, Sound Effects at 200 a generation, Music at 900 a minute and Dubbing from 2,000, all from one shared monthly pool. T1Official Other

Voice realism leads every review

0/2

All seven open on it. Litmus scores the platform 89 of 100 and ToolProven calls it the voice quality benchmark. Other

A documented developer path

0/2

The API, its documentation and its SDKs, named in passing by three of the seven. Other

Cheap to start, with cloning early

0/2

Instant Voice Cloning is listed from Starter, and two review sites point at the $6 tier as the affordable way in. T1Official Other

App Store headline ratings

Both listings rate the platform highly on the day we looked. T3Buyers

The stores' own headline figures, one request each on 19 September 2026. Not reviews we read.

What keeps coming up against it

Billed regenerations at four review sites; credits that expire on cancellation at two, and in the company's own FAQ.

Failed and regenerated takes are billed

Well documented, size unknown
Documented byReview sitesT1ElevenLabsHow commonfour of the seven review sites we read describe their own runs, and none of them counts how common it is
Review sites

Four describe paying for takes they did not use. Litmus puts the real cost at about 2.8 times the advertised figure, AIwithKay at two to three times what the plan page implies, and AI Alleyway and Nerdynav each report regenerations eating credits.

Litmus · read 19 SepAIwithKay · read 19 SepAI Alleyway · read 19 SepNerdynav · updated 1 Aug

T1ElevenLabs

Its pricing FAQ says credits are charged per generation request rather than per download, and that a limited number of free regenerations may be available while the content and settings do not change.

ElevenLabs · read 19 Sep

T3Buyers

Not counted. No buyer review was read one by one, and the two store figures are headline ratings only. Thin

Credits expire when you cancel

0/2

Two review sites raise it, and ElevenLabs' own FAQ agrees: upgrading rolls unused credits into the next cycle, while downgrading or cancelling lets unused paid credits expire. T1Official Other

Non-English output and pronunciation slip

0/2

Two of the seven say so, both writing from English-language work. Other

Where they split

Three places the evidence pulls two ways, starting with the blind-listener boards.

The arena against the hands-on reviews

T2Editorial

The blind-listener boards on 19 September 2026 put Eleven v3 fifth on controlled voices and fourteenth on providers' own voices, behind Cartesia, Inworld, Alibaba and Speechify models.

Other

Against that: All seven review sites we read call it the best in the category or the benchmark for it.

Both trueThe boards compare one voice of one model against clones of the same eight voices. The review sites describe a product with a catalogue, two cloning routes, an editor and an API around it.

Buyers against the billing complaints

T3Buyers

4.84 from 20,527 ratings and 4.72 from 10,278, read on 19 September 2026.

Other

Against that: One review site reports the platform at 3.2 out of 5 on Trustpilot over the credit rules, and another puts real cost near 2.8 times the plans.

Our readThe ratings answer whether the voices are worth using. The complaints answer whether the bill matches the page.

The cheaper models rank lower

T1Official

Flash v2.5 is listed at fifty per cent lower price per character for API generations and around 75 milliseconds of latency.

T2Editorial

Against that: The same board ranks Flash v2.5 and Turbo v2.5 near its bottom.

Our readSpeed and price are built for a different job from expressive delivery, and the board scores only one of those jobs.

Who it's for, who should pass

For one subscription over the whole audio pipeline. Pass if you need a bill that matches the plan page.

It suits you if

  • You want one account for voice, transcription, dubbing, sound effects and music, paid from a single credit pool.Its pricing FAQ says every product draws on one shared monthly pool, and six of the seven review sites mention the breadth.
  • You are wiring speech into software and want documented clients and a low-latency model.Three of the seven review sites name the API, docs and SDKs; Flash v2.5 is listed at around 75 milliseconds.
  • You want to hear the quality before paying.The free tier carries 10,000 credits a month.
  • You are localising content and can check your own language first.Two of the seven review sites report non-English output slipping.

Pass if

  • You need a bill that matches the plan page.Four of the seven review sites report paying for takes they did not use.
  • Your volume is high and your margin per minute is thin.Litmus says this is not the cheapest per minute at scale.
  • You need the top of a blind-listener arena.Three other vendors' models sit above Eleven v3 on both boards.
  • You need terms that stay put.Its Terms of Use, last updated 31 March 2026, reserve the right to change limits, replace or suspend services and change fees.

This is a reading of published reviews, not medical, financial or legal advice. For a decision about your health, your money or your rights, a qualified professional is the right next step, and not a review.

The verdict

The breadth is real and under-reported. The credit rules are the recurring argument.
VerdictProvisionalReview sites in, owners uncounted

One subscription that covers voice, transcription, dubbing, sound effects and music for a platform whose voice quality every review site we read leads with, and whose credit rules four of the seven report billing above the plan page.

Confidence, by tier

Officialstrong

Six company pages read 19 September 2026 carry the plans, models, credit costs and the published cloning safeguards.

Editorialthin

Two blind-listener boards, strong on rank and price, and no review of the product as a whole in this tier.

Buyersthin

Two headline ratings. No buyer review was read one by one.

Forumthin

One API total of 409 stories; the thread read carried no product finding.

Otherthin

Seven review sites and creator blogs, four dated and three not, several carrying affiliate links.

Editorial evidence
under a month old
Newest report
None read
Read
19 Sep · month 12

Share line

One subscription, the whole audio pipeline. The meter runs on the takes you deleted.
Review MachineElevenLabs · Master Review

Rests onElevenLabs' pricing FAQ charges credits per generation request rather than per download, and four of the seven review sites we read report paying for takes they did not use.

Sources

18 sources, heaviest tier first. Every figure above comes from one of these.

  1. T1OfficialProhibited Use Policyupdated 3 Sep 2025
  2. T1OfficialTerms of Useupdated 31 Mar
  3. T1OfficialElevenLabs pricingread 19 Sep
  4. T1OfficialElevenLabs modelsread 19 Sep
  5. T1OfficialVoice cloning docsread 19 Sep
  6. T1OfficialSafetyread 19 Sep
  7. T2EditorialControlled voice leaderboardread 19 Sep
  8. T2EditorialProvider voice leaderboardread 19 Sep
  9. T3BuyersElevenLabs: AI Voice Generator headline ratingheadline rating and count onlyread 19 Sep
  10. T3BuyersElevenReader headline ratingheadline rating and count onlyread 19 Sep
  11. ForumSearch API total for ElevenLabsone request reporting a totalread 19 Sep
  12. OtherAIDigitalBox review29 Apr
  13. OtherAmrytt review27 Jun
  14. OtherToolProven reviewupdated 19 Jul
  15. OtherNerdynav reviewupdated 1 Aug
  16. OtherLitmus reviewno date of its ownread 19 Sep
  17. OtherAIwithKay reviewno date of its ownread 19 Sep
  18. OtherAI Alleyway reviewno date of its ownread 19 Sep

MethodOn 19 September 2026 we read six ElevenLabs pages, two blind-listener leaderboards, seven review sites and creator blogs from 29 April to 1 August 2026 or undated, two App Store headline ratings and one public Hacker News figure, by web search and plain page fetch, and synthesised them with AI; no model was prompted, no plan was bought and nothing was tested.

Dates without a year are 2026.

The review as text

The same review as one piece of writing · 7 min read

ElevenLabs sells AI audio on monthly credits: Free at $0 with 10,000 credits, Starter at $6 with 30,000, Creator at $22 with 121,000, Pro at $99 with 600,000, Scale at $299 with 1.8 million and Business at $990 with 6 million, with Enterprise priced on request. Its models page lists Eleven v3 and Eleven v3 Conversational for expressive speech, Multilingual v2 for long-form reads, Flash v2.5 and Flash v2 for speed, Scribe v2 and Scribe v2 Realtime for transcription, and Eleven Music v2. This is a reading of the platform as its pages stood on 19 September 2026.

The sample: six of the company's own pages, two blind-listener leaderboards, seven review sites and creator blogs, four dated between 29 April and 1 August 2026 and three carrying no date of their own, two App Store headline ratings and one public Hacker News figure. No established publication's scored review of the platform could be read, so the counts below rest on named, dated review sites and blogs and on ElevenLabs' own pages.

Consensus

Voice quality is the one thing nobody in the sample argues about. All seven review sites lead with it: Litmus scores the platform 89 of 100 and calls it its pick, ToolProven puts it top on voice quality, and the other five open on the same claim. The Artificial Analysis boards, which rank models by blind listener votes, put Eleven v3 fifth on controlled voices at an Elo of 1048 and fourteenth on providers' own voices at 1166, with Cartesia, Inworld, Alibaba and Speechify models ahead and Cartesia's Sonic 3.6 first on 1117.

So the reading is strong on realism and thin everywhere else. The seven review sites judge a product: a voice library, two cloning routes, an editor and an API around the model. We read the two kinds of evidence as answering two questions, and neither answers the other.

On cloning, ElevenLabs publishes its own position and we report it as that: its docs say Professional Voice Cloning is limited to your own voice and needs a verification step even where the owner consents, its Prohibited Use Policy bars unauthorised or deceptive impersonation, and its safety page says generated audio can be traced to the account that made it. No source we counted describes what gets past those safeguards.

Recurring strengths

One account covers the whole audio pipeline. This is the benefit that recurs in passing and almost never leads: six of the seven review sites mention in a clause that the same subscription also does transcription, dubbing, sound effects, music and voice agents. Litmus lists a full audio pipeline in one account among its pros, AIwithKay says one account covers text to speech, voice cloning, dubbing, music, transcription and conversational agents, and AI Alleyway, Amrytt, ToolProven and Nerdynav each describe the same breadth in a line before returning to voice quality. ElevenLabs' pricing FAQ gives the mechanics: one shared monthly credit pool, with Speech to Text at 330 credits a minute, Sound Effects at 200 a generation, Music at 900 a minute and Dubbing from 2,000.

Voice realism with directed delivery. Seven of the seven review sites. ToolProven timed a first take of 991 characters on Multilingual v2 at 12.5 seconds.

A documented developer path. Three of the seven review sites name the API, its docs and its SDKs: Litmus, AI Alleyway and Amrytt.

Cheap to start, and cloning comes early. Nerdynav and ToolProven both point at the $6 Starter tier, which lists Instant Voice Cloning.

Recurring complaints

Failed and regenerated takes are billed. Four of the seven review sites report it. Litmus puts the real cost at about 2.8 times the advertised figure, AIwithKay at two to three times what the plan page implies, and AI Alleyway and Nerdynav each describe regenerations eating credits. ElevenLabs' pricing FAQ confirms the mechanic without the arithmetic: credits are charged per generation request rather than per download, and free regenerations are limited and conditional. Nobody we read counts how common this is.

Credits expire when you cancel. Two review sites raise it, AIwithKay and AI Alleyway, and ElevenLabs' fee page agrees: upgrading rolls unused credits into the next cycle, while downgrading or cancelling lets unused paid credits expire.

Non-English output and pronunciation slip. Two of the seven say so, Litmus and Nerdynav, both writing from English-language work.

Where reviewers split

The arena against the hands-on reviews. The boards put three other vendors' models above Eleven v3 on both boards on 19 September 2026, while all seven review sites call it the best in the category or the benchmark for it. We read this as a difference of job: winning a paired comparison between voices, against being worth the subscription around them.

Buyers against the billing complaints. The App Store headline for the voice generator reads 4.84 out of 5 from 20,527 ratings, and ElevenReader 4.72 from 10,278. Against that, AI Alleyway reports the platform at 3.2 out of 5 on Trustpilot over the credit rules, and Litmus puts real cost near 2.8 times the plans. The ratings answer whether the voices are worth using; the complaints answer whether the bill matches the page.

The cheaper models rank lower. The boards price Eleven v3 and Multilingual v2 at $100 per million characters and v3 Conversational and Flash v2.5 at $50, with Cartesia's Sonic 3.6 at $49. ElevenLabs lists Flash v2.5 at fifty per cent lower price per character for API generations and around 75 milliseconds of latency, and the same board ranks Flash v2.5 and Turbo v2.5 near its bottom. Speed and price are built for a different job, and the board scores only one.

Who it suits

You want one account for voice, transcription, dubbing, sound effects and music, paid from a single credit pool. You are wiring speech into software and want documented clients and a low-latency model. You want to hear the quality before paying: the free tier carries 10,000 credits a month. You are localising content and can check your own language first: two of the seven review sites report non-English output slipping.

Who should pass

You need a bill that matches the plan page, because four of the seven review sites report paying for takes they did not use. Your volume is high and your margin per minute is thin: Litmus says this is not the cheapest per minute at scale. You need the top of a blind-listener arena, where three other vendors' models sit above Eleven v3. You need terms that stay put: ElevenLabs' Terms of Use, last updated 31 March 2026, reserve the right to change limits, replace or suspend services and change fees.

This is a reading of published reviews, not medical, financial or legal advice. For a decision about your health, your money or your rights, a qualified professional is the right next step, and not a review.

One verdict a week: the most useful Master Review we finished, the complaint that kept appearing, and who should skip it. The weekly verdict is here.

Sources

On 19 September 2026 we read six ElevenLabs pages, two blind-listener leaderboards, seven review sites and creator blogs from 29 April to 1 August 2026 or undated, two App Store headline ratings and one public Hacker News figure, by web search and plain page fetch, and synthesised them with AI; no model was prompted, no plan was bought and nothing was tested.

One verdict a week.

Every week, the most useful Master Review we finished: what the internet agrees on, the complaint that kept appearing, and who should skip it.

By subscribing you agree to receive one weekly email from Review Machine. You can unsubscribe at any time.