LTX-2 Fast is a high-speed video generation model developed by Lightricks as part of the LTX-2 foundation model family. Optimized for rapid ideation, the model is designed to generate short-form video content with synchronized audio. It serves as an efficiency-focused variant within the LTX-2 ecosystem, prioritizing low-latency inference and responsiveness for workflows such as storyboarding and quick prototyping.
The model is built on a Diffusion Transformer (DiT) architecture and utilizes a multiscale rendering approach. This design allows it to generate video frames significantly faster than standard diffusion models while maintaining subject consistency. A standout capability of LTX-2 Fast is its native support for synchronized audio-video generation, where motion and sound are produced in a single pass to ensure temporal alignment without the need for external audio models.
LTX-2 Fast supports both text-to-video and image-to-video generation, producing cinematic motion with features such as camera pans, zooms, and dolly shots. It is capable of generating clips up to 20 seconds long at 1080p resolution. As part of the open-source LTX-2 release, the model weights are available for local execution and are optimized to run on consumer-grade GPUs.