AURA: Always-On Understanding and Real-Time Assistance via Video Streams

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lu, Xudong, Bo, Yang, Chen, Jinpeng, Li, Shuhan, Guo, Xintong, Guan, Huankang, Liu, Fang, Xu, Dunyuan, Sun, Peiwen, Sun, Heyang, Liu, Rui, Li, Hongsheng
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915917482426368
author Lu, Xudong
Bo, Yang
Chen, Jinpeng
Li, Shuhan
Guo, Xintong
Guan, Huankang
Liu, Fang
Xu, Dunyuan
Sun, Peiwen
Sun, Heyang
Liu, Rui
Li, Hongsheng
author_facet Lu, Xudong
Bo, Yang
Chen, Jinpeng
Li, Shuhan
Guo, Xintong
Guan, Huankang
Liu, Fang
Xu, Dunyuan
Sun, Peiwen
Sun, Heyang
Liu, Rui
Li, Hongsheng
contents Video Large Language Models (VideoLLMs) have achieved strong performance on many video understanding tasks, but most existing systems remain offline and are not well-suited for live video streams that require continuous observation and timely response. Recent streaming VideoLLMs have made progress, yet current approaches often rely on decoupled trigger-response pipelines or are limited to captioning-style narration, reducing their effectiveness for open-ended question answering and long-horizon interaction. We propose AURA (Always-On Understanding and Real-Time Assistance), an end-to-end streaming visual interaction framework that enables a unified VideoLLM to continuously process video streams and support both real-time question answering and proactive responses. AURA integrates context management, data construction, training objectives, and deployment optimization for stable long-horizon streaming interaction. It achieves state-of-the-art performance on streaming benchmarks and supports a real-time demo system with ASR and TTS running at 2 FPS on two 80G accelerators. We release the AURA model together with a real-time inference framework to facilitate future research.
format Preprint
id arxiv_https___arxiv_org_abs_2604_04184
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle AURA: Always-On Understanding and Real-Time Assistance via Video Streams
Lu, Xudong
Bo, Yang
Chen, Jinpeng
Li, Shuhan
Guo, Xintong
Guan, Huankang
Liu, Fang
Xu, Dunyuan
Sun, Peiwen
Sun, Heyang
Liu, Rui
Li, Hongsheng
Computer Vision and Pattern Recognition
Video Large Language Models (VideoLLMs) have achieved strong performance on many video understanding tasks, but most existing systems remain offline and are not well-suited for live video streams that require continuous observation and timely response. Recent streaming VideoLLMs have made progress, yet current approaches often rely on decoupled trigger-response pipelines or are limited to captioning-style narration, reducing their effectiveness for open-ended question answering and long-horizon interaction. We propose AURA (Always-On Understanding and Real-Time Assistance), an end-to-end streaming visual interaction framework that enables a unified VideoLLM to continuously process video streams and support both real-time question answering and proactive responses. AURA integrates context management, data construction, training objectives, and deployment optimization for stable long-horizon streaming interaction. It achieves state-of-the-art performance on streaming benchmarks and supports a real-time demo system with ASR and TTS running at 2 FPS on two 80G accelerators. We release the AURA model together with a real-time inference framework to facilitate future research.
title AURA: Always-On Understanding and Real-Time Assistance via Video Streams
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.04184