Deepgram Launches Flux TTS: An 80ms Voice Model Built for Live Conversation

deepgramscott · x · 2026-08-12

Deepgram has launched Flux TTS, a new text-to-speech model specifically designed for real-time voice conversations.

Core Capabilities:

Ultra-Low Latency: Responds in as little as 80ms, enabling a lifelike conversational experience.
Context-Aware: Tracks context across turns and adapts dynamically mid-call.
Complex Scenarios: Optimized for real-world customer service challenges like handling interruptions, changed orders, and fast string-of-numbers readouts.

Developers can build and test the model for free until September 12th.

More from Apps