The "answer only when spoken to" era of audio AI is over. Meet a model that keeps listening to sound, environment, and instructions — and acts on its own 🎧 Title: Audio Interaction Model URL: 🎧 Overview A unified, always-on streaming Large Audio Language Model (LALM). It runs a continuous perceive-decide-respond loop, listening to sound, environmental audio, and user instructions at once, then responding dynamically based on semantic understanding of the stream. ❓ Challenges Solved Most current LALMs run offline and handle isolated tasks (streaming ASR or voice chat separately). Real interaction needs an always-on model that listens in real time and reacts on the fly. 💡 Methodology & Proposed Approach The core is the SoundFlow framework operationalizing the perceive-decide-respond loop. ・Streaming-native data construction ・Comprehension-aware training ・Asynchronous low-latency inference for stable real-time interaction It trains on StreamAudio-2M (2.6M items) covering 7 fundamental abilities and 28 sub-tasks, plus Proactive-Sound-Bench to assess proactive intervention. 📊 Experimental Results / Use Cases ・Maintains competitive performance across 8 benchmarks ・Enables capabilities offline LALMs can't: real-time ASR, streaming audio instruction following, and proactive intervention Great for always-on voice assistants, real-time dialogue, and proactive audio assistance. #AudioAI# #LALM#

With cultural heritage as our oars and the spirit of our times as our vessel, we invite you to join us for a #DragonBoatFestival# audiovisual feast where ancient charm meets modern trends, and warmth blends with depth! Stay tuned on June 18th, #2026AdventuresonDragonBoatFestival#

With cultural heritage as our oars and the spirit of our times as our vessel, we invite you to join us for a #DragonBoatFestival# audiovisual feast where ancient charm meets modern trends, and warmth blends with depth! Stay tuned on June 18th, #2026AdventuresonDragonBoatFestival#

Congratulations to @nuance_ai on the $50M Series A led by @lightspeedvp, with @Accel, @spc, @nvidia, and @definevc joining the round. What makes Nuance Labs compelling is that the team is working on a problem most AI products still avoid: human conversation is about much more than words. Tone, timing, gaze, hesitation, facial expression, and the simple act of showing that you are listening all carry meaning. The founding team has unusually deep experience for this challenge. Fangchang Ma, Edward Zhang, and Karren Yang are former Apple researchers with PhDs spanning robotics, machine learning, computer graphics, and audiovisual synthesis. They have spent years working on how machines perceive, reconstruct, and respond to people. Their approach is also refreshingly ambitious. Instead of stitching together transcription, an LLM, voice generation, and facial animation, Nuance Labs is building one full duplex audiovisual model that can see, hear, reason, speak, and express itself in real time. That distinction matters. Most AI avatars still pause awkwardly, interrupt at the wrong moment, or stare blankly while someone is speaking. Nuance is treating responsiveness and active listening as part of the model itself, not as polish added later. There is a huge opportunity here across coaching, education, sales, customer service, training, and any setting where trust and communication affect the outcome. As AI becomes more capable, raw intelligence will not be enough. The products people actually want to spend time with will also need presence, timing, and emotional awareness. Nuance Labs has the technical depth, product conviction, and patience to take on this hard problem properly. Excited to see the public research preview later this year. This is one of the teams pushing human AI interaction in a genuinely important direction.