🤖 NVIDIA has introduced Audex — a unified audio-text model based on the Nemotron 3 Nano architecture.
The 30B version supports a context of up to 1 million tokens and up to 15 minutes of audio, enabling simultaneous sound understanding, speech generation, and textual reasoning.
🌍 The emergence of full Unified Audio Language Models (UALM) narrows the gap between ASR and TTS, allowing for the construction of systems with complex logic and deep understanding of audio contexts.
👤 This is a step toward creating voice assistants that truly "hear" and understand audio nuances while maintaining the level of intellectual reasoning found in text-based LLMs.
Source 1: https://huggingface.co/nvidia/Nemotron-Labs-Audex-30B-A3B/tree/main
Source 2: https://huggingface.co/nvidia/Nemotron-Labs-Audex-30B-A3B/blob/main/MODELCARD.md