PersonaPlex

#16 in KI-Inferenz-Hardware

nvidia · 3× · tolest 30. Juni 2026

21

Momentum

PersonaPlex is a 7B speech-to-speech model by NVIDIA based on Moshi. It enables real-time, full-duplex communication where listening and speaking occur simultaneously. Voice and persona can be controlled through text prompts and audio conditioning.

Belege (3)