- ElevenLabs released Scribe v3 for pre-recorded audio and Scribe v3 Live for real-time transcription.
- Both models are built to handle messy audio conditions like crosstalk and background noise.
- Independent benchmarks from Artificial Analysis found Scribe v3 among the most accurate transcription models, but slower and more expensive than rivals from Microsoft and xAI.
- Scribe v3 Live uses conversational context to avoid cutting off callers during brief pauses.
What Happened
ElevenLabs released two new speech recognition models, Scribe v3 and Scribe v3 Live, built for messy audio conditions such as overlapping speakers in meetings or background noise on phone calls, The Decoder reported on November 12, 2026. Scribe v3 targets pre-recorded files, while Scribe v3 Live is built for low-latency real-time use, and both are available now.
Why It Matters
Voice agent reliability has become a competitive battleground as companies push AI assistants into phone-based customer service, where rivals like xAI’s Grok STT and TTS APIs have also targeted phone-call accuracy. ElevenLabs is betting that handling crosstalk and background noise correctly will differentiate its models in enterprise deployments.
Technical Details
Independent benchmarks from Artificial Analysis found Scribe v3 among the most accurate transcription models on the market, narrowly beating rivals from OpenAI, Google, and Microsoft, but the model runs slower and costs more than competing options from Microsoft and xAI. Scribe v3 Live’s key technical feature is analyzing conversational context in real time to avoid cutting off callers during brief pauses, a common failure mode in earlier voice-agent transcription systems.
Who’s Affected
ElevenLabs is expanding enterprise partnerships with Nvidia and McKinsey alongside the launch, positioning the models for large-scale enterprise voice-agent deployments rather than just consumer transcription tools. Developers building call-center and meeting-transcription products now have another accuracy-focused but higher-cost option to weigh against cheaper rivals.
What’s Next
Buyers will need to weigh Scribe v3’s accuracy edge against its higher latency and cost, particularly for high-volume, real-time deployments where Microsoft and xAI’s cheaper models remain viable alternatives.
