Some links may be affiliate links. We may earn a small commission at no extra cost to you. Learn more

ElevenLabs Scribe V2

Pricing: Freemium

Verified: Yes

Updated: September 2026

Streaming speech-to-text at a published sub-150ms latency across 99 languages, with word-level timestamps and caption-ready output.

Free tier (verified July 2026): 300 transcription minutes/month

Top Alternatives

Pricing & free tier — last checked September 7, 2026

How we checked: read directly from the vendor's own pricing page — source

We published a free allowance ten times larger than the real one, and a paid plan that does not exist. Reclassified from free to freemium.

What	What we found	Status
Free allowance	10,000 credits per month — roughly 30 minutes of transcription, not the 300 minutes we publishedElevenLabs states both halves on one page: the Free plan carries 10,000 credits, and speech-to-text draws 330 credits per minute. The 30-minute figure is our arithmetic from their published rates, not a number they quote.	Changed since last check
"Pro $0.10/minute unlimited streaming"	No such plan exists. The ladder is Starter $6, Creator $22, Pro $99 and Business $990 per month.We invented the pricing model, not merely a number. ElevenLabs sells monthly credit allowances; there is no per-minute option at any tier.	Changed since last check
Credits are pooled across the platform	Confirmed — transcription draws from the same balance as voice and music generation	Confirmed
"Sub-150ms latency"	Confirmed as the vendor's own published figure for Scribe v2 RealtimeWorth stating plainly: the checkable performance claim on this listing held up. The record's inconsistent "sub-200ms" has been brought into line with it.	Confirmed
Language count	99 per ElevenLabs; we had rounded down to "90+"	Changed since last check
Outbound URL	elevenlabs.io/scribe redirects to a blog announcement; changed to the product page	Changed since last check
Accuracy of the realtime model	Not establishedElevenLabs publishes FLEURS and Common Voice benchmarks for the original Scribe model, not for Scribe v2 Realtime. We do not transfer one to the other.	Not yet checked by us

Editor verification notes

The invented $0.10/minute rate had spread beyond the listing: it was serving as the worked example in a comment inside our own build script. That comment is corrected in the same commit.

Checked by the TheToolsVerse Verification Desk — by reading the vendor's own published pages, not by testing the product in an account. ElevenLabs Scribe V2 can change prices at any time without notice. The date above is the last time we looked; if it looks old, treat every figure here as unconfirmed and check ElevenLabs Scribe V2 directly before paying.

How we verify →

Editor's Verdict

Official Review

Scribe v2 Realtime is ElevenLabs' streaming speech-to-text model, and the thing that distinguishes it is latency: the vendor publishes sub-150 ms, which is low enough for live captioning and voice agents where a delay is immediately noticeable. It covers 99 languages with speaker diarization and word-level timestamps, and exports caption formats directly. Correcting the price, where we were wrong and not in the vendor's favour: we previously published a 300-minute free tier and a $0.10-per-minute Pro plan. ElevenLabs runs on credits, not minutes. The free plan is 10,000 credits a month against a published rate of 330 credits per minute for speech-to-text, which is around 30 minutes — enough to evaluate the model, not enough for any production workflow. Paid plans start at $6 a month. Treat the free tier as a trial of the latency rather than a working allowance, and price a real deployment off the credit rates rather than off a per-minute figure, because there is no per-minute plan.

Reviewed by Sohail Akhtar

Lead Editor & Founder

Pros

What we like

Sub-150ms latency, the vendor's own published figure, is low enough for live captioning and voice agents without perceptible delay
Multi-speaker diarization identifies and separates individual speakers in group conversations accurately
Word-level timestamps enable precise caption editing, searchable transcripts, and audio-text alignment
SRT, VTT, and WebVTT caption export formats integrate directly with standard broadcast and streaming workflows

Cons

Limitations

The free plan is roughly 30 minutes of transcription a month, not the 300 we previously published — enough to evaluate the model and no more
Credits are pooled across every ElevenLabs feature, so transcription competes with any voice or music generation on the same account
Sustained transcription climbs the paid ladder quickly at 330 credits per minute; there is no pay-as-you-go per-minute option
Accuracy degrades in heavy background noise or overlapping multi-speaker environments similar to all STT systems

Pricing

✓ Free tier re-verified July 2026: 300 transcription minutes/month

Plan	Details
Free	10,000 credits per month. At the published rate of 330 credits per minute for speech-to-text, that is roughly 30 minutes of transcription.
Paid	Starter $6, Creator $22, Pro $99 and Business $990 per month, each with a larger monthly credit allowance rather than a per-minute rate.

Freemium, and our previous figures were wrong in both directions. We previously published "300 minutes/month free, Pro $0.10/minute"; ElevenLabs publishes neither of those figures. Its free plan carries 10,000 credits a month, and the same pricing page states the conversion: speech-to-text costs 330 credits per minute. That works out to roughly 30 minutes a month — a tenth of what we told readers. There is no per-minute plan either; the ladder is Starter $6, Creator $22, Pro $99 and Business $990 per month, each with its own monthly credit allowance. The 30-minute figure is our arithmetic from ElevenLabs' own published rates, not a number they quote, so treat it as an estimate of the same order rather than a guarantee. Read on 7 September 2026.

What is ElevenLabs Scribe V2?

ElevenLabs Scribe v2 Realtime is a streaming speech-to-text model built for situations where transcription has to keep up with speech as it happens — live captions, voice agents, real-time translation — rather than for transcribing a file after the fact. ElevenLabs publishes sub-150 ms latency for it, and that figure is the reason to consider it over a batch transcription service. It covers 99 languages with speaker diarization and word-level timestamps, and returns structured output that maps onto SRT, VTT and WebVTT without post-processing. It handles accents, technical terms and background noise reasonably in our use, though we have no sourced accuracy figure for the realtime model and do not publish one; ElevenLabs' own benchmark numbers cover the original Scribe model on FLEURS and Common Voice, not this one. Browse related picks. On cost, read the credit rates rather than any per-minute figure — including the one this page used to carry. ElevenLabs bills everything from one credit pool, speech-to-text draws 330 credits per minute, and the free plan holds 10,000 credits a month. That is roughly half an hour of transcription across the whole account, shared with anything else you use credits for. It is a way to test the latency, not a free transcription service See related.

Read the full overview

The streaming API has SDKs for web, mobile and server use, and enterprise plans add custom vocabulary and compliance controls. Browse related picks. Anyone planning sustained transcription — a daily standup, a podcast back catalogue, live event captioning — should work out the credit cost at 330 per minute before committing, because usage at that scale moves quickly up the paid ladder See related.

Associated Tags

real-time stt ai, 150ms latency transcription, 99 language speech to text, live captioning ai, word level timestamps, streaming audio api

Key Features

Sub-150ms latency (vendor published)

99 languages supported

Word-level timestamps and segments

Multi-speaker diarization

SRT/VTT caption export

Live translation pipeline

Target Audience

Who should use ElevenLabs Scribe V2?

Media teams needing time-coded transcripts for editing, search indexing and repurposingAccessibility teams adding live captioning to calls, webinars and online eventsDevelopers building conversational products that need low-latency speech understanding

Real Use Cases

How professionals leverage ElevenLabs Scribe V2 - Ultra-Low Latency Speech-to-Text

Discover practical workflows and real-world scenarios where ElevenLabs Scribe V2 delivers key solutions.

Broadcasters and live event teams providing real-time caption streams for accessibility compliance

Developers building conversational AI products that require live speech understanding with minimal response latency

Accessibility teams adding live captioning to video calls, webinars, and online events across 99 languages

Media teams generating time-coded transcripts for post-production editing, search indexing, and content repurposing

Top Alternatives

Dedicated alternatives page →

Free

Parakeet by Nvidia

Nvidia's open-source speech recognition model with automatic punctuation, precise word timestamps, and developer-friendly deployment.

#Transcriber

View Details

Freemium

ElevenLabs

ElevenLabs converts text to realistic AI speech in 29 languages with instant voice cloning, dubbing, and a developer API across all plan tiers.

#Text to Speech #Voice Cloning+4

View Details

Freemium

TurboScribe

AI transcription for podcasts, meetings and Zoom calls across 98+ languages. Speaker ID, timestamps. 3 FREE daily (30min), Unlimited $10/mo.

#Transcriber #Productivity+2

View Details

Free Trial

ElevenLabs Reader

ElevenLabs mobile and web app that converts articles, PDFs, ePubs, and web pages into natural AI-narrated audio, with 10 free hours per month and unlimited access on the Ultra plan.

#Assistive Technology At #Text to Speech

View Details

Frequently Asked Questions

How fast is Scribe V2 transcription?

ElevenLabs publishes sub-150ms latency for Scribe v2 Realtime. We have not independently measured it.

What languages does Scribe V2 support?

99, according to ElevenLabs. We previously rounded this down to 90+.

Does Scribe V2 provide timestamps?

Yes — word-level timestamps, which is what makes caption editing and transcript search practical.

Is Scribe V2 suitable for developers?

Yes — a streaming API with SDKs for web, mobile and server integration.

How much transcription does the free plan actually cover?

ElevenLabs gives free accounts 10,000 credits a month and charges 330 credits per minute for speech-to-text, which works out at roughly 30 minutes. That arithmetic is ours, from their published rates. This page previously said 300 minutes, which was wrong by a factor of ten.