Story

On air. The free wire runs 15 minutes behind live. See pricing

World, crypto, finance and tech news, scored for impact.

CryptoTokenPostKorean, translatedFiled 13 h ago

9 billion parameter Aura 2.0, watch the video and start a conversation first

Original headline90억 매개변수 아우라 2.0, 영상 보고 먼저 대화 시작

Open the sourceReported by TokenPost, tokenpost.kr, 19 November 2026, by 최윤서

Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.

Huawei Research, CUHK MMLab, and Hong Kong University of Science and Technology researchers released AURA 2.0, a 9-billion-parameter AI model that proactively initiates conversation by selecting speakable moments in video without waiting for user prompts. The model combines real-time video question answering with external tool calling and on-screen object referencing, built around three core capabilities: model-initiated conversation (MIC), video-context-based tool calling (SATC), and interactive visual grounding (IVG). The team detailed the system in a technical report submitted on the 13th (Korea time), with model scale listed on its Hugging Face model card. The release matters because it shifts video AI from passive query-response toward agentic, context-aware interaction that decides when to speak and which tools to invoke. What to watch next is broader benchmarking and adoption of the open model, and whether proactive video agents of this size become practical for real-time applications.

Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.

Named in this story

AURA 2.0, Huawei Research, Hong Kong University of Science and Technology, CUHK MMLab, Hugging Face