CogVideo

Developed by Tsinghua University, the CogVideoX series is an open-source model family that supports tasks from text to video, image to video, and video continuity. HuggingFace works integrated with Diffusers and CogKit frameworks, allowing for fine-tuning (fine-tuning) and inference processes with low VRAM consumption. The series provides models of different parameter sizes, such as 5B and 1.5-5B, allowing video production in various hardware configurations.