Lightricks releases LTX-2.5 with image-to-video task

Lightricks has published the LTX-2.5 model on HuggingFace for the image-to-video task. The publication is dated 07.23.2026, the model has 39 downloads and 124 likes. The code is released as open source, details are on the model page.

What it means

LTX-2.5 solves the image-to-video task, that is, it animates a still image. In our reference on AI for 3D and video, Lightricks tools are not yet featured, and the closest in spirit are 3D generator pages with the image-to-3d task: OpenLRM, Hunyuan3D 2.1 and TRELLIS. That is a different class of tasks: they turn an image into a three-dimensional object rather than a video clip, so LTX-2.5 does not directly overlap with them. Nevertheless, the arrival of an open source model from Lightricks expands the landscape of tools that take an image as input and produce a moving result. For users choosing between generating a 3D asset and animation, this is one more option in an adjacent niche. The developer does not specify the technical requirements or the size of the model weights for LTX-2.5 on the model page. It is too early to say whether LTX-2.5 will earn a place in the CPU3D reference: that would require data on licenses, hardware requirements and output formats, none of which are in the publication.