VideoCrafter

VideoCrafter generates video on your own hardware from written prompts or still images. It's aimed at researchers and people exploring video generation for personal use; the toolbox limits use to personal, research and non-commercial purposes.

• Text-to-video generation: Describe a scene's subjects, actions and style to generate clips, including realistic scenes, painterly animation and low-poly imagery.
• Image animation: Start with a still image and use a text prompt to describe the movement you want in the clip.
• Model choice: VideoCrafter2 provides text-to-video generation focused on visual quality, motion and scene composition, while VideoCrafter1 includes both text-to-video and image-to-video models.
• Local browser interface: The Gradio demo lets you try the text and image workflows in a browser while inference runs on your machine.

Pretrained checkpoints are available through Hugging Face, so generating clips doesn't require training a model yourself. The toolbox uses Python. The toolbox README limits use to personal, research and noncommercial purposes. Check the terms attached to each checkpoint before use.

The toolbox provides VideoCrafter1 and VideoCrafter2 checkpoints and source installation. Its README restricts toolbox use to personal, research and noncommercial purposes. The VideoCrafter2 model page labels its checkpoint Apache 2.0, but that label should not be read as resolving the toolbox’s broader usage notice.