LiveCC Conversation and Real-Time Commentary - Gradio Demo

LiveCC: Learning Video LLM with Streaming Speech Transcription at Scale (CVPR 2025)

1️⃣ Select Mode, Real-Time Commentary (LiveCC) or Conversation (Common QA/Multi-turn)

2️⃣🅰️ Real-Time Commentary: Input a query (optional) -> Click or upload a video.

2️⃣🅱️ Conversation: Click or upload a video -> Input a query. But as the past_key_values support in ZeroGPU is not good, multi-turn conversation could be slower.

HF Space Gradio has unsolvable latency (10s~20s), and not support flash-attn. If you want to enjoy the very real-time experience, please deploy locally https://github.com/showlab/livecc