LTX launches new free-to-use open world model for video and physical AI - roboticsandautomationnews.com

LTX has released LTX-2.5, an open-weights world model for video generation and physical AI. The model is described as free to use, and it is distributed through the Hugging Face platform. It includes native support for ComfyUI. The announcement states that LTX-2.5 offers improvements in visual quality, prompt understanding, generation speed, and efficiency. The release is aimed at users working on both video generation and physical AI.

The 'world model' designation indicates that the system is intended to generate or reason about physical scenes, rather than only producing individual images or clips. In the context of physical AI, such models can serve as a simulation or planning component for robots and other embodied systems. The open-weights approach means the model's parameters are publicly released, allowing developers to run it locally or fine-tune it. Hosting on Hugging Face provides a single repository for downloading and sharing the model. Native ComfyUI support is listed as a feature of the release.

The release adds to a growing set of open-weights models that target not just media generation but also embodied AI use cases. By combining video generation with physical AI in one free package, LTX is offering a tool that can be applied across animation, synthetic data creation, and robot training. The intended audience appears to include researchers, hobbyists, and commercial developers. The announcement does not include further technical details, such as parameter counts or training data. It frames the improvements as making the model more practical for real-world applications.