A research unit with a model lineage
LTX describes itself as a lab for video and world models that began as a division of Lightricks. Its own account identifies the team that built LTX-Video and LTX-2 from scratch, including the diffusion architecture, video autoencoder and inference stack. This establishes a research unit and its work, beyond the use of LTX as a model name. [1]
Lightricks' current page says the model and licensing business is still in the process of separating. The lab's origins and continuing research are established; that wording does not establish a completed corporate transfer. The current account does not identify finalized post-separation ownership. [2]
Models as an infrastructure component
The May 2025 LTXV-13B release is part of that lineage. Lightricks published the model with multiscale rendering and developer-accessible artifacts, making video generation a component that other software could run and adapt instead of only a finished editing app. [3][1]
The original 13B open-weights terms included a revenue threshold for commercial licensing and restrictions on use. Those terms concern that released model; the lab's broader ambition to build models for physical-world reasoning is a research direction, not proof that every output accurately simulates physics. [4][1]
Sources
- About Us
LTX · retrieved Oct 23, 2026
- About Lightricks
Lightricks · retrieved Oct 23, 2026
- Lightricks Launches 13B Parameters LTX Video Model, Breakthrough Rendering Approach Generates High-quality, Efficient AI Video 30X Faster Than Comparable Models
Lightricks · May 6, 2025
- Upload ltxv-13b-0.9.7-dev-fp8.license.txt with huggingface_hub
Lightricks · May 6, 2025