ByteDance today launched SeedRealtime, a native audio-video full-duplex model that integrates audio, video, and text into a unified architecture for real-time multimodal interaction. Available on the Doubao app, the model is designed for simultaneous seeing, listening, and speaking with minimal latency.