Google DeepMind has released Genie 4, its first real-time world model featuring native audio generation for ambient sounds, music, and character dialogue. The system allows users to dynamically rewrite simulations via text commands (e.g., changing weather or camera angles), coordinated by Gemini 3.8 Flash acting as an in-loop director. It has been validated in simulating humanoid robot tasks and is currently available only to trusted testers.
Real-time Inference
实时推理
Ability of a model to process input and generate output within millisecond-level latency
Curated Intelligence
The iRTE2026 conference highlights the shift in AI voice interaction from single-turn Q&A to full-duplex continuous communication. Agora, as an infrastructure provider, showcased its underlying capabilities supporting models like OpenAI and Gemini, aiming to lower development barriers by reusing network and interaction components. The event primarily reflects industry trends and vendor marketing, lacking specific technical breakthroughs or parameter details.
Overworld has open-sourced Waypoint-1 on Hugging Face, a video diffusion model designed for real-time interactive generation. The release includes weights for the 'Small' variant to enable low-latency experiences, accompanied by a hackathon to foster ecosystem development.
OpenAI officially launched GPT-4o, a new flagship model capable of real-time reasoning across audio, vision, and text. This release marks a significant leap from pure text processing to native multimodal real-time interaction.