Lightricks, the company behind the Facetune and Videoleap apps, has released a new video model called LTX Video, or LTXV, that generates what the company describes as high-quality AI video at speeds up to 30 times faster than competing products, and does it using consumer-grade hardware. The open-source, 13-billion parameter model achieves such efficiency by utilizing an approach called multiscale rendering, which generates video in progressively detailed layers. The program can run on high-end laptops and standard desktop computers, opening up generative video to an audience beyond those who have access to enterprise equipment.

VentureBeat notes that “a major challenge for AI video generation has been the enormous computational requirements,” with popular models from companies including Runway, Pika Labs and Luma AI typically running on cloud configurations that use “multiple enterprise-grade GPUs with 80GB or more of VRAM (video memory), making local deployment impractical for most users.”

Lightricks co-founder and CEO Zeev Farbman characterizes the debut of LTX Video as “a pivotal moment in AI video generation” that will allow more people to “create content with more consistency, better quality, and tighter control.”

Inspired by fine artists who sometimes begin complex projects with rough sketches then add the finer points, multiscale rendering “allows the model to generate details gradually,” Farbman tells VentureBeat, explaining that “you’re starting on the coarse grid, getting a rough approximation of the scene, of the motion of the objects moving, etc. And then the scene is kind of divided into tiles, and every tile is filled with progressively more details.”

LTXV also requires less memory by tapping compressed latent space, Farbman added in VB. “The advantage for AI is that ‘your peak amount of VRAM is limited by a tile size, not the final resolution.’”

There, “it can be licensed freely by any organization with less than $10 million in annual revenue,” which means “users are free to tinker with it any way they want, fine-tune it, add new features and integrate it into third-party applications,” writes SiliconANGLE.

The original LTXV model debuted in late 2024 with 2 billion parameters, getting “a lot of attention as one of the most advanced video generation models around” with a lightweight architecture that allowed it to “run efficiently on laptops and personal computers powered by a single consumer-grade graphics processor, rapidly generating five seconds of slick-looking video with smooth and consistent motion,” according to SiliconANGLE.

Additional details on LTXV are available via the press release and on GitHub and Hugging Face.