LTX-Video: Video Generator — What It Does and What You Need to Run It

LTX-Video is an open source video generator from Lightricks that solves the image-to-video task. The tool turns a still image into a video clip using running on your own hardware.

What it does

The authors position the model as a solution for image-to-video generation. The repository tags also mention text-to-video and image-to-video-generation tasks, which indicates support for text descriptions as an additional condition. It is built on diffusion-models and dit architectures, the code is written in Python and integrated with the diffusers library.

What you need to run it

The code is open under the Apache-2.0 license, the model weights are distributed under the LTXV Open Weights License 0.X. The largest weights file takes up 26.6 GB, and the total size of all weights files is 236.4 GB.

As described by the author, running it requires 1 GB of VRAM — verbatim: "Requires only 1GB of VRAM". MPS support on macOS with PyTorch 2.3.0 is claimed. The codebase was tested with Python 3.10.5, CUDA 12.2 and supports PyTorch versions 2.1.2 and higher. The repository was created in November 2024, the last code change dates to January 2026.

Who it suits

The tool is aimed at those who are ready for local deployment and have enough disk space to store the weights. The stated 1 GB VRAM requirement makes the model accessible to a wide range of GPUs, including relatively old or low-power cards. MPS support allows it to run on Mac computers with Apple Silicon.

It is not suitable for those looking for a cloud service without local installation — LTX-Video requires self-deployment from the repository. Users without experience with Python and the command line may find the entry barrier high.

The model and code are available on GitHub and Hugging Face, more information can be found on the project website. The model weights are distributed under the LTXV Open Weights License 0.X: commercial use is limited by a revenue threshold.