VideoCrafter: video generator — what it does and what you need to run it

VideoCrafter is a set of open-source models for generating video from a text description or a starting image. The tool was created by the Tencent AI Lab and targets researchers and developers who want to experiment with video diffusion models while having access to the source code.

What it does

• Text-to-video generation. The model creates a short video based on a text description.
• Image-to-video generation. You can feed a static image as input and get an animated scene.
• Working with diffusion models. The authors describe VideoCrafter2 as a solution aimed at overcoming data limitations for high-quality video diffusion models.

What you need to run it

• Platform. The code is written in Python. Running it requires an environment with Python support and the relevant libraries.
• Hardware. The authors do not specify exact minimum requirements for VRAM or system RAM. Since these are diffusion models, running locally will require a discrete GPU with a sufficient amount of VRAM.
• License. The code is open source under the Apache License 2.0 with a restriction: non-commercial use only.
• Repository. The source code is available at github.com/AILab-CVC/VideoCrafter. The last code change is dated January 9, 2026.

Who it suits

VideoCrafter is aimed primarily at researchers and developers familiar with the Python ecosystem and diffusion models. The tool will be useful to those who need open source code for fine-tuning, modifying, or studying the architecture of generative video models. It is not intended for out-of-the-box consumer use without programming skills. Also note that commercial use, including selling, distributing, and incorporating it into commercial products or services, is strictly prohibited.

A project page with additional information is available at ailab-cvc.github.io/videocrafter2/. The latest release at the time of the repository analysis is VideoCrafter1, while the authors describe work on a second version.