ElevenLabs has released two speech-to-text models, Scribe v3 and Scribe v3 Live. Scribe v3 is built for pre-recorded audio files. Scribe v3 Live targets low-latency, real-time applications. Both are available through the ElevenAPI developer interface and the ElevenAgents platform.

According to ElevenLabs, Scribe v3 cuts word errors by about 25 percent compared with its predecessor across 102 languages. The company says German accuracy improves significantly. It also says speaker diarization works more reliably in larger group conversations, and that Scribe v3 Live offers this capability in real time for the first time.

Independent benchmarks from Artificial Analysis rank Scribe v3 among the most accurate transcription models available. It edges out models from Microsoft, Mistral, xAI, Google, and OpenAI, and trails only a preview model from Alibaba. The same benchmarks show that ElevenLabs is mid-pack on speed and price. Models from Microsoft and xAI run many times faster at marginally higher error rates and cost roughly half as much. The benchmark measures only word errors on recorded files, so it does not test the company's claims about speaker diarization, rare langu