Recently, the DeepBeepMeep team released Wan2.1GP on GitHub, a video generation model optimized for users with low-end GPUs. Based on Alibaba's Wan2.1, this model aims to provide powerful video generation capabilities for users lacking high-performance GPU resources. The launch of Wan2.1GP marks a significant advancement in video generation technology, especially within the open-source domain.
Image Source Note: Image generated by AI, licensed by Midjourney.
Wan2.1GP's key features include its excellent performance and broad applicability. The model consistently outperforms existing open-source models and some commercial solutions in multiple benchmark tests, demonstrating strong competitiveness. Furthermore, the T2V-1.3B model requires only 8.19GB of VRAM, making it runnable on almost all consumer-grade GPUs. Using an RTX 4090, users can generate a 5-second 480P video in approximately 4 minutes, achieving performance comparable to some closed-source models.
Wan2.1GP not only supports various tasks such as text-to-video, image-to-video, and video editing, but it's also the first model capable of generating videos with both Chinese and English text. This feature opens up more possibilities for practical applications. Additionally, the model incorporates a powerful video variational autoencoder (VAE), enabling efficient encoding and decoding of 1080P videos of any length while preserving temporal information, laying a solid foundation for video and image generation.
To enhance user experience, Wan2.1GP has undergone several optimizations, including significantly reduced memory and VRAM requirements, and support for various configurations to accommodate devices with different performance levels. Users can quickly get started with this tool through a simplified installation process. With continuous updates, Wan2.1GP is gradually incorporating more practical features, such as Tea Cache support and Gradio interface improvements, further increasing generation speed and ease of use.
Project Link: https://github.com/deepbeepmeep/Wan2GP
Highlights:
👍 SOTA Performance: Wan2.1GP excels in multiple benchmark tests, surpassing existing open-source and commercial solutions.
🖥️ High Compatibility: Requires only 8.19GB of VRAM, supporting almost all consumer-grade GPUs, ideal for low-end users.
📜 Multi-Task Support: Supports various generation tasks including text-to-video and image-to-video, and features both Chinese and English text generation capabilities.