LiveStar: An AI Co-Host for Live Streams
Meet LiveStar — an AI co‑host for live streams that understands what’s happening on screen and speaks up at the right moment.
Most video AIs do great offline, but stumble live: they process frames slowly and interrupt or fall behind. LiveStar fixes that with:
- Adaptive streaming decoding: keeps a coherent narrative as the scene evolves, frame by frame.
- “Response‑silence” timing: a one‑pass check that decides when to talk — and when to stay quiet.
- Memory‑aware acceleration: compresses past context to handle 10+ minute videos; plus a streaming cache for 1.53× faster inference.
Results: state‑of‑the‑art across three benchmarks — +19.5% semantic correctness, 18.1% better timing, and +12% FPS on all five OmniStar tasks.
They also built OmniStar, a training and evaluation set spanning 15 real‑world scenarios and 5 online‑video tasks.
Paper: http://arxiv.org/abs/2511.05299v1 • Code & data: https://github.com/yzy-bupt/LiveStar
Paper: http://arxiv.org/abs/2511.05299v1
Register: https://www.AiFeta.com
AI VideoAI LiveStreaming LLM ComputerVision Streaming Research OpenSource Dataset
