ByteDance’s Seed team has released the end-to-end simultaneous interpretation model Seed LiveInterpret 2.0

ByteDance’s Seed team has launched Seed LiveInterpret 2.0, an end-to-end simultaneous interpretation model that supports bidirectional Chinese-English translation. It delivers near-human translation accuracy with ultra-low latency (as low as 3 seconds), enabling real-time “listen and speak” translation even for multi-speaker audio input. Its zero-shot voice cloning feature can instantly replicate the speaker’s voice without requiring any pre-collected samples.