Source
Build Low-Latency Multilingual Voice Agents: Open Weights & Full Deployment Control with NVIDIA Magpie TTS
NVIDIA has released Magpie Multilingual TTS, an open-weights text-to-speech model designed for building low-latency voice agents across 12 languages.
Why it matters
Allowing deployment on private infrastructure reduces response delays and improves data privacy for enterprises building real-time voice applications like healthcare assistants.
The details
- Supports 12 languages, including newly added Korean, Arabic, and Brazilian Portuguese.
- Achieves 32ms Time to First Audio on single streams using B200 GPUs.
- Uses frame stacking and local transformers to increase audio generation speed.
Get the weekly recap
The stories like this one, picked and explained — once a week, straight to your inbox.