Seed_realtime

Excited to release SeedRealtime, a native audio-visual full-duplex LLM. As a core contributor, I worked on visual-based proactive interaction and duplex RL training, enabling the model to continuously watch, listen, and respond at the right moment.