LongCat-VideoLongCat-Video
Open Source🚀 Based on Meituan LongCat-Video 13.6B - MIT License!

LongCat-Video: Next-Generation AI Video Generator Model

LongCat-Video: an open-source AI that creates videos from text or images, and continues existing videos (Text-to-Video, Image-to-Video, and Video-Continuation tasks).

What is LongCat-Video

LongCat-Video is a foundational video generation model from Meituan with 13.6B parameters. It features a unified architecture supporting Text-to-Video, Image-to-Video, and Video-Continuation tasks within a single model. LongCat-Video natively supports long video generation, producing minutes-long videos without color drifting or quality degradation, and generates 720p 30fps videos efficiently using coarse-to-fine generation strategy.

  • Text-to-Video Generation
    Transform text prompts into high-quality videos with advanced neural networks. Create engaging visual content from simple text descriptions.
  • Image-to-Video Generation
    Bring static images to life by generating smooth video animations. Create dynamic content from single images with intelligent motion synthesis.
  • Video Continuation
    Extend existing videos seamlessly without quality loss or color drifting. Generate minutes-long videos with consistent quality and natural transitions.

Key Features of LongCat-Video

Advanced AI-powered video generation capabilities designed for creators and developers worldwide.

Unified Multi-Task Framework

Single model supporting Text-to-Video, Image-to-Video, and Video-Continuation tasks with consistent high performance across all generation modes.

Long Video Generation

Natively pretrained on Video-Continuation tasks to produce minutes-long videos without color drifting or quality degradation.

High-Quality Output

Generate 720p, 30fps videos with professional quality using advanced neural network architecture and optimization techniques.

Efficient Inference

Fast video generation within minutes using coarse-to-fine generation strategy along temporal and spatial axes with GPU acceleration.

Strong Performance

Powered by multi-reward GRPO optimization achieving performance comparable to leading open-source and commercial video generation models.

Open Source & MIT License

Fully open-source model with MIT license available on GitHub and Hugging Face for research and commercial applications.

What People Are Talking About LongCat-Video on X

Join the conversation about LongCat-Video on social media

FAQ