Deepgram Launches Flux Multilingual Conversational Speech Recognition Model

Deepgram, a real-time AI infrastructure provider for voice applications, introduced Flux Multilingual, a conversational speech recognition model that supports 10 languages and can automatically detect, understand and switch languages within a single conversation in real time.

The model extends beyond traditional transcription-focused automatic speech recognition by enabling natural dialogue flow with accurate turn-taking, interruption handling and low latency. Supported languages are English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian and Dutch.

Scott Stephenson, Deepgram CEO and co-founder, said Flux Multilingual advances the ability for global enterprises to interact with customers via voice agents, noting the model can switch languages mid-call to maintain a seamless experience.

Omar Paul, Twilio VP of products, said the model lets teams extend English conversational experiences across languages without stitching together multiple systems or sacrificing accuracy.

Key features include model-based turn detection with end-of-turn decisions in under 400 milliseconds, monolingual-grade accuracy across languages, automatic language detection with optional hints, native code-switching, mid-session configurability and compatibility with existing Flux API integrations. The offering is available via Deepgram’s Cloud API or as a self-hosted deployment, with EU endpoints and SDK support.

Deepgram said Flux Multilingual is generally available and is being offered at a limited-time promotional rate for streaming speech-to-text, including Flux Multilingual and Nova-3 models. Developers can learn more at https://deepgram.com or try Flux Multilingual in the Deepgram Playground.

The post Deepgram Launches Flux Multilingual Conversational Speech Recognition Model appeared first on TV News Check.