Case Study: Telnyx achieves ultra-low-latency real-time voice AI with Deepgram Flux

Telnyx delivers sub-second voice AI with Deepgram Flux

Telnyx, a global communications and connectivity platform, faced the challenge of unpredictable latency and unstable performance from third-party speech AI APIs, which made voice AI conversations feel mechanical and queued. To power truly real-time voice experiences at carrier scale, Telnyx needed a solution that could process speech directly within its own network perimeter without forcing audio off-net to external providers.

By embedding the Deepgram Flux speech-to-text engine directly into its media plane and running it on Telnyx-managed GPUs at its global Points of Presence, Telnyx achieved ultra-low-latency, natural conversations. This Deepgram solution enabled stable performance during regional call spikes, reliable barge-in, and a simpler operational model, resulting in more responsive voice AI experiences that feel instant to end-users.

Telnyx