ModelsPublished All news from that day
ElevenLabs releases Scribe v3 Live, which it reports cuts semantic word error rate 43% and word error rate 31%.
Scribe v3 transcribes recorded audio and v3 Live handles realtime audio, with speaker labels in realtime for the first time. The accuracy figures are ElevenLabs' own, and the pricing page lists $0.20 and $0.22 per hour.
Check our sources · 12 facts from 2 sourcesElevenLabs released Scribe v3, for recorded audio, and Scribe v3 Live, for realtime audio, on 10 November.
It says v3 makes 27% fewer word errors than Scribe v2 across 102 languages, and that v3 Live cuts semantic word error rate by 43% and plain word error rate by 31% against Scribe v2 Realtime. v3 Live also adds speaker labels, called diarization, to realtime transcription for the first time.
ElevenLabs announced both in a blog post on 10 November.
ElevenLabs frames them as the transcript step for products such as meeting notes, voice agents, dubbing and clinical documentation. It says v3 makes 17% fewer diarization errors than v2 on recordings with three or more speakers, and that v3 Live adds turn detection that uses the conversation's context, not only silence, so a caller who pauses mid-number is not cut off.
All the comparisons are ElevenLabs' own. The multilingual figures come from an internal version of the FLEURS benchmark, and the diarization tests use a dataset ElevenLabs built from real meetings, agent calls and voice notes in 16 languages, which it says it will open-source soon.
ElevenLabs also reports that Eneco, an energy supplier in the Netherlands, tested v3 Live in its voice agent and saw Dutch postcodes recognised correctly on the first try 40% more often, and house numbers 15% more often.
v3 Live is available in ElevenAgents, both models are in the ElevenAPI, and both are the default in the ElevenLabs app. The API pricing page lists Scribe v3 at $0.20 per hour of audio and Scribe v3 Live at $0.22 per hour, the same as Scribe v2 and v2 Realtime. ElevenLabs says its Zero Retention Mode deletes audio and text as soon as a request completes.
Underlined sentences link to the facts we checked. Dotted words open a short explainer.
Fact 6 of 12 · from ElevenLabs, 10 Nov 2026
Multilingual accuracy was measured on an internal version of the FLEURS benchmark ... we also built our own dataset of real meetings, agent calls, and voice notes across 16 languages. We’ll be open sourcing this soon.
Your reaction
We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.
We count reactions per story and day, never who reacted. They help choose what goes in the monthly issue.
Check our sources
Every sentence above is checked against these 2 sources.
1 Introducing Scribe v3 and v3 Live, our most accurate speech to text models
Open the source archived copy-
ElevenLabs' blog post of November 10, 2026 introduces Scribe v3, for recorded audio, and Scribe v3 Live, the low-latency variant for realtime audio.
citeToday ElevenLabs is launching two new Speech to Text models built for exactly these conditions:
-
ElevenLabs says Scribe v3 reduces word error rate by 27% compared with Scribe v2 across 102 languages on its internal version of FLEURS, and Scribe v3 Live reduces semantic WER by 43% (and WER by 31%) compared with Scribe v2 Realtime.
Toned down to what the source says
citeScribe v3 Live ... makes 43% fewer errors than Scribe v2 Realtime. ... Across 102 languages, Scribe v3 Live reduces semantic WER by 43% and WER by 31% compared with Scribe v2 Realtime.
-
ElevenLabs says Scribe v3 Live includes speaker diarization (labeling who said what) for the first time, and improved turn detection.
citeIt also includes speaker diarization for the first time (labeling who said what) and improved turn detection.
-
ElevenLabs says Scribe v3 makes 17% fewer diarization errors than Scribe v2 on recordings with three or more speakers.
citeScribe v3 makes 17% fewer diarization errors than Scribe v2 on recordings with three or more speakers
-
ElevenLabs says Scribe v3 Live uses the context of the conversation as well as the audio for turn detection, so a caller who pauses halfway through a number isn't cut off.
citeScribe v3 Live uses the context of the conversation as well as the audio, so a caller who pauses halfway through a number isn’t cut off.
-
ElevenLabs says it measured multilingual accuracy on an internal version of the FLEURS benchmark, and built its own diarization dataset of real meetings, agent calls, and voice notes across 16 languages, which it says it will open-source soon.
citeMultilingual accuracy was measured on an internal version of the FLEURS benchmark ... we also built our own dataset of real meetings, agent calls, and voice notes across 16 languages. We’ll be open sourcing this soon.
-
ElevenLabs says Eneco, which supplies energy to more than 2 million homes and businesses in the Netherlands, tested Scribe v3 Live on its ElevenAgents voice agent, and that in real customer calls the agent recognized Dutch postcodes correctly on the first try 40% more often, and house numbers 15% more often.
citeEneco, which supplies energy to more than 2 million homes and businesses in the Netherlands, runs its voice agent on ElevenAgents and tested Scribe v3 Live on it ahead of launch. In real customer calls, the agent recognized Dutch postcodes correctly on the first try 40% more often, and house numbers 15% more often.
-
ElevenLabs says Scribe v3 Live is available in ElevenAgents, both models are available in ElevenAPI, and both are the default in the ElevenLabs app.
citeScribe v3 Live is available in ElevenAgents, and both v3 and v3 Live are the default in the ElevenLabs app. ... Both models are also available in ElevenAPI today.
-
ElevenLabs says Zero Retention Mode deletes audio and text as soon as a request completes, and that the models run in the US by default with isolated EU, India and Singapore environments for enterprise customers with data residency requirements.
citeScribe v3 and Scribe v3 Live run in the US by default, and are offered in isolated EU, India, and Singapore environments for enterprise customers with data residency requirements. Zero Retention Mode deletes audio and text as soon as a request completes.
-
ElevenLabs' blog post says Scribe v3 reduces word error rate by 27% compared with Scribe v2 across 102 languages on FLEURS, and that results measured in different ways aren't directly comparable.
citeAcross 102 languages on FLEURS, Scribe v3 reduces WER by 27% compared with Scribe v2 and improves in every language tested. ... results measured in different ways aren’t directly comparable.
-
ElevenLabs says products built on audio start with a transcript, such as meeting notes, voice agents, dubbing, clinical documentation and broadcast archives, and that Scribe v3 and v3 Live are built for the messy conditions of real-world audio.
citeProducts built on audio start with a transcript: meeting notes, voice agents, dubbing, clinical documentation, and broadcast archives that span decades of content. ... Today ElevenLabs is launching two new Speech to Text models built for exactly these conditions:
2 ElevenLabs API pricing
Open the source archived copy-
ElevenLabs' API pricing page lists Scribe v3 speech to text at $0.20 per hour and Scribe v3 Live at $0.22 per hour, the same per-hour prices it lists for Scribe v2 ($0.20) and Scribe v2 Realtime ($0.22).
cite
Every sentence above is checked against these 2 sources. Each fact has its own address you can cite.
Source 1 · ElevenLabs · 10 Nov 2026
Introducing Scribe v3 and v3 Live, our most accurate speech to text models
ElevenLabs' blog post of November 10, 2026 introduces Scribe v3, for recorded audio, and Scribe v3 Live, the low-latency variant for realtime audio.
Today ElevenLabs is launching two new Speech to Text models built for exactly these conditions:
ElevenLabs says Scribe v3 reduces word error rate by 27% compared with Scribe v2 across 102 languages on its internal version of FLEURS, and Scribe v3 Live reduces semantic WER by 43% (and WER by 31%) compared with Scribe v2 Realtime.
Toned down to what the source saysScribe v3 Live ... makes 43% fewer errors than Scribe v2 Realtime. ... Across 102 languages, Scribe v3 Live reduces semantic WER by 43% and WER by 31% compared with Scribe v2 Realtime.
ElevenLabs says Scribe v3 Live includes speaker diarization (labeling who said what) for the first time, and improved turn detection.
It also includes speaker diarization for the first time (labeling who said what) and improved turn detection.
Show all 11 factsShow fewer facts
ElevenLabs says Scribe v3 makes 17% fewer diarization errors than Scribe v2 on recordings with three or more speakers.
Scribe v3 makes 17% fewer diarization errors than Scribe v2 on recordings with three or more speakers
ElevenLabs says Scribe v3 Live uses the context of the conversation as well as the audio for turn detection, so a caller who pauses halfway through a number isn't cut off.
Scribe v3 Live uses the context of the conversation as well as the audio, so a caller who pauses halfway through a number isn’t cut off.
ElevenLabs says it measured multilingual accuracy on an internal version of the FLEURS benchmark, and built its own diarization dataset of real meetings, agent calls, and voice notes across 16 languages, which it says it will open-source soon.
Multilingual accuracy was measured on an internal version of the FLEURS benchmark ... we also built our own dataset of real meetings, agent calls, and voice notes across 16 languages. We’ll be open sourcing this soon.
ElevenLabs says Eneco, which supplies energy to more than 2 million homes and businesses in the Netherlands, tested Scribe v3 Live on its ElevenAgents voice agent, and that in real customer calls the agent recognized Dutch postcodes correctly on the first try 40% more often, and house numbers 15% more often.
Eneco, which supplies energy to more than 2 million homes and businesses in the Netherlands, runs its voice agent on ElevenAgents and tested Scribe v3 Live on it ahead of launch. In real customer calls, the agent recognized Dutch postcodes correctly on the first try 40% more often, and house numbers 15% more often.
ElevenLabs says Scribe v3 Live is available in ElevenAgents, both models are available in ElevenAPI, and both are the default in the ElevenLabs app.
Scribe v3 Live is available in ElevenAgents, and both v3 and v3 Live are the default in the ElevenLabs app. ... Both models are also available in ElevenAPI today.
ElevenLabs says Zero Retention Mode deletes audio and text as soon as a request completes, and that the models run in the US by default with isolated EU, India and Singapore environments for enterprise customers with data residency requirements.
Scribe v3 and Scribe v3 Live run in the US by default, and are offered in isolated EU, India, and Singapore environments for enterprise customers with data residency requirements. Zero Retention Mode deletes audio and text as soon as a request completes.
ElevenLabs' blog post says Scribe v3 reduces word error rate by 27% compared with Scribe v2 across 102 languages on FLEURS, and that results measured in different ways aren't directly comparable.
Across 102 languages on FLEURS, Scribe v3 reduces WER by 27% compared with Scribe v2 and improves in every language tested. ... results measured in different ways aren’t directly comparable.
ElevenLabs says products built on audio start with a transcript, such as meeting notes, voice agents, dubbing, clinical documentation and broadcast archives, and that Scribe v3 and v3 Live are built for the messy conditions of real-world audio.
Products built on audio start with a transcript: meeting notes, voice agents, dubbing, clinical documentation, and broadcast archives that span decades of content. ... Today ElevenLabs is launching two new Speech to Text models built for exactly these conditions:
Source 2 · ElevenLabs
ElevenLabs API pricing
ElevenLabs' API pricing page lists Scribe v3 speech to text at $0.20 per hour and Scribe v3 Live at $0.22 per hour, the same per-hour prices it lists for Scribe v2 ($0.20) and Scribe v2 Realtime ($0.22).
Topics
The morning email
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep