Introducing Fluent: Next-Generation Multilingual Transcription for Voice Agents

Fluent is a new multilingual speech-to-text model introduced on the Bland platform, designed for real-time, two-way voice conversations with enhanced transcription accuracy and speed. Fluent significantly reduces transcription errors with a word error rate (WER) of approximately 5.9% in English, outperforming leading competitors and baselines like OpenAI's Whisper. The model features improved end-of-speech detection, reducing interruptions and latency, and handles intra-utterance language switching for seamless transcription of bilingual conversations. Currently supporting English, Spanish, German, French, Portuguese, and Italian, Fluent is optimized for precision in these common enterprise languages, whereas other models like Auto and Babel offer broader language coverage. Fluent's deployment is straightforward through the API or dashboard, providing a seamless fallback to Auto in case of issues, and is recommended for agents operating in English, Spanish, or Western European languages to enhance conversation quality.

No tracked trend matches for this post yet.