On August 5, 2026, ByteDance officially released SeedRealtime, a native full-duplex audio-visual model with a unified architecture that natively integrates audio, video, and text. It supports real-time interaction within a continuous multimodal information stream, enabling simultaneous viewing, listening, and speaking. Compared to traditional cascaded systems, SeedRealtime reduces latency and information loss, enhancing the natural flow of conversation. Currently, the model has been fully rolled out on the Doubao App.