Shypd.aiThe largest open index of AI toolsPusa-VidGen is a video generation tool that leverages vectorized timestep adaptation for fine-grained temporal control. It offers fast 4-step inference for various video tasks, including text-to-video and image-to-video.
Trending
About
Pusa-VidGen is a groundbreaking video diffusion model that introduces a paradigm shift through frame-level noise control with vectorized timesteps. This approach enables fine-grained temporal control within a unified video diffusion framework. By finetuning the SOTA Wan-T2V-14B model with vectorized timestep adaptations (VTA), Pusa V1.0 achieves unprecedented efficiency, surpassing Wan-I2V on Vbench-I2V with significantly lower training costs and data. It supports both Wan2.1 and Wan2.2 architectures, with the latter leveraging a MoE DiT architecture for enhanced quality. The integration of LightX2V allows for lightning-fast 4-step inference while maintaining high quality, making it suitable for various tasks like text-to-video, image-to-video, video extension, and video transitions.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
Also listed in
