🚀 UniVA: Universal Video Agents towards Next-Generation Video Intelligence

univa-agent is a revolutionary video agent designed to provide an unprecedented interactive experience and high-quality video generation capabilities.

Our goal is not just to release a model, but to build a unified and powerful video creation platform.

🔗 Project Ecosystem

univa-agent is a complete open-source project, including the following resources:

🚀 Try the Demo (Invitation-Only)

We provide an online Demo to quickly experience the powerful features of univa-agent.

Please note: Demo access is currently by [Invitation-Only]. We are committed to providing a stable, high-quality experience for our initial users.

• ➡️ Click here to access the Demo
• ➡️ Apply for Beta Access

Core Features

The design of univa-agent is built on two core pillars, aiming to simultaneously address both the generation quality and the creative experience of video content.

🌟 Pillar 1: Unprecedented Interactive Experience

We believe the future of video creation should be interactive and intelligent. univa-agent introduces:

• Unified Interaction System: Handle multiple video tasks within a single, unified framework without needing to switch tools.
• Agent with Memory: Capable of understanding multi-turn conversational context to perform complex, stateful video editing and creation.
• Deep Interaction Capabilities: Supports fine-grained instructions, enabling comprehensive control from high-level concepts down to specific details.

🎨 Pillar 2: High-Quality Generation Capabilities

A powerful agent must be matched with high-quality execution. univa-agent ensures:

• Broad Task Support (Breadth): Covers a wide range of functions, from text-to-video generation, video editing, and style transfer to video inpainting.
• High-Fidelity Video Output: Generated video content achieves state-of-the-art results in clarity, coherence, and visual aesthetics.
• Powerful Function Map: [Briefly describe 1-2 unique functional modules mentioned in your meeting, e.g., "Synergistic Components" or "Architecture Highlights"].

👥 Team

This project is developed by the UniVA team. For detailed team member introductions, please visit our Team Page.

✏️ How to Cite

If you find our work helpful for your research, please consider citing our paper:

@misc{liang2025univauniversalvideoagent,
      title={UniVA: Universal Video Agent towards Open-Source Next-Generation Video Generalist}, 
      author={Zhengyang Liang and Daoan Zhang and Huichi Zhou and Rui Huang and Bobo Li and Yuechen Zhang and Shengqiong Wu and Xiaohan Wang and Jiebo Luo and Lizi Liao and Hao Fei},
      year={2025},
      eprint={2511.08521},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2511.08521}, 
}