# Kling Releases UniVideo for Generation and Understanding

> The new open-source model combines both video generation and comprehension, a rare dual capability built on the Qwen2.5 vision-language foundation.

Published by The Open Weights on Oct 18, 2025. Canonical: https://theopenweights.com/news/univideo-5yrf

## Key facts

- Company: Kuaishou
- Model: UniVideo
- Version: 1.0
- Category: Any-to-Any
- Modalities: Any-to-Any, Text → Video
- License: Apache 2.0 (Open weights, commercial use allowed)
- Significance: notable
- Published: 2025-10-18
- Last verified: 2026-06-17
- Hugging Face: https://huggingface.co/KlingTeam/UniVideo
- Canonical URL: https://theopenweights.com/news/univideo-5yrf

The Kling team has released UniVideo, a new open-source model designed to both generate and understand video content. Unlike many models that focus solely on text-to-video synthesis, UniVideo operates as a unified system, capable of interpreting the contents of a video as well as creating new ones from text prompts.

At its core, UniVideo is built upon Qwen2.5-VL-7B, a powerful large vision-language model. This foundation provides a strong base for processing and relating visual and textual information, allowing a single model architecture to handle tasks that often require separate, specialized systems. This unified approach can lead to more efficient and coherent video processing.

## Why It Matters

While the open-source community has made significant strides in video generation, models that also possess deep comprehension abilities are less common. UniVideo helps bridge this gap by providing a single, powerful tool for more complex video-related AI tasks. By combining generation with understanding, it enables new possibilities for content analysis, automated description, and creative workflows within a single framework.

The model is released under a permissive Apache 2.0 license, encouraging broad adoption and experimentation. Researchers and developers can access the model and its code on its [Hugging Face repository](https://huggingface.co/KlingTeam/UniVideo) to explore its dual capabilities.

## Get the model

- [Hugging Face](https://huggingface.co/KlingTeam/UniVideo)

## Sources

- [KlingTeam/UniVideo](https://huggingface.co/KlingTeam/UniVideo) — Hugging Face, Oct 18, 2025

---
Source: The Open Weights (https://theopenweights.com/). Aggregated and written by Claude, curated by humans. Cite as: "Kling Releases UniVideo for Generation and Understanding", The Open Weights, Oct 18, 2025, https://theopenweights.com/news/univideo-5yrf