**[ByteDance Releases SeedRealtime, a Full-Duplex Audio-Video Model]** According to a report by CLS, ByteDance has officially launched its native full-duplex audio-video model, SeedRealtime. SeedRealtime integrates audio, video, and text through a unified architecture, enabling real-time interaction across continuous multimodal information streams, delivering a new "watch, listen, and speak simultaneously" experience. End-to-end human evaluation results show that compared to cascade models, SeedRealtime reduces audio-video dialogue rhythm issues by half—demonstrating a more natural grasp of speaking timing. Issues such as "interrupting before the speaker finishes, delayed responses after the speaker finishes, or being mistakenly triggered by background noise or casual chatter" have been significantly reduced. Additionally, the probability of completing a single conversation smoothly and coherently has also seen a notable improvement. Currently, SeedRealtime has been fully rolled out on the Doubao App.