Vibitely Daily

Friday, 13 November 2026, story 3 of 7

ElevenLabs releases Scribe v3 and v3 Live speech-to-text models

Artificial Analysis ranks Scribe v3 near the top on accuracy, but Microsoft and xAI models run many times faster at about half the price.

Illustration: A balance scale with a large, finely cut crystal on one pan and a stack of small, plain, fast-looking arrow-shaped tokens on the other, tipping neither way.

What happened

ElevenLabs has released two speech-to-text models, Scribe v3 for pre-recorded audio and Scribe v3 Live for real-time use. The Decoder reports both are available now through the ElevenAPI developer interface and the ElevenAgents platform. The source gives no prices.

ElevenLabs says Scribe v3 cuts word errors by about 25 percent against its predecessor across 102 languages, and that speaker diarization is more reliable in larger groups. It says Scribe v3 Live brings diarization to real time for the first time. Enterprise customers get EU-isolated servers and a zero-retention mode, the company says.

What it means

For voice-agent builders, the choice is no longer simply the most accurate model. Artificial Analysis puts Scribe v3 behind only an Alibaba preview model on word errors, yet the source says Microsoft and xAI models are many times faster at marginally higher error rates and cost roughly half as much. In a live call, speed may matter more than a small accuracy gain.

The benchmark covers only word errors on recorded files. It does not test diarization, rare languages or the real-time behavior of Scribe v3 Live, which is where ElevenLabs makes its main claims. The context-based end-of-speech detection is so far the company's description, not independently measured. Teams that lose callers to cut-offs, or handle noisy and overlapping audio, have the clearest reason to test it. Others may find the cost and speed gap decisive.

What to do: If your voice agent cuts callers off or struggles with crosstalk, test Scribe v3 Live on your own audio and compare latency and cost against faster rivals.

Source

Filed under Models. Published .

"What it means" is analysis, not reporting. Found an error? Write to daily@vibitely.com.