Lightricks Launches Fast AI Video Generation Model for Consumer GPUs

Lightricks, the company behind popular creative apps like Facetune and VideoLeap, has announced the release of its most powerful AI video generation model yet: the LTX Video 13-billion-parameter model (LTXV-13B). This model is capable of generating high-quality AI videos up to 30 times faster than comparable models, and uniquely, it runs efficiently on consumer-grade GPUs rather than requiring expensive enterprise hardware.

The key innovation behind LTXV-13B is its novel multiscale rendering technique, which generates video progressively in layers of detail. This approach mirrors how artists create complex scenes by starting with rough sketches and adding finer details over time. By dividing the video frame into tiles and rendering details incrementally, the model drastically reduces peak VRAM usage, allowing it to run on GPUs like Nvidia’s 3090, 4090, and 5090, including laptop variants.

This efficiency breakthrough addresses one of the biggest challenges in AI video generation: the enormous computational and memory requirements that typically restrict models to cloud-based enterprise GPUs with 80GB or more VRAM. Lightricks’ model compresses the latent space and limits memory usage per tile, enabling professional-quality video generation on standard desktop and laptop hardware.

Beyond technical innovation, Lightricks has chosen to open-source LTXV-13B, making it available on platforms like Hugging Face and GitHub. This open approach encourages academic research, community-driven improvements, and wider adoption, contrasting with many AI companies that keep their models behind closed APIs. Open sourcing also helps reduce R&D costs and fosters collaboration across the AI ecosystem.

To address growing legal concerns around AI training data, Lightricks has partnered with Getty Images and Shutterstock to license content for model training. This strategy reduces copyright risks, making LTXV-13B more attractive for enterprise use where compliance and legal clarity are critical.

In a strategic move to foster innovation, Lightricks offers free licensing of LTXV-13B to startups with under $10 million in annual revenue. This encourages early adoption and community building, with plans to negotiate licensing fees for larger enterprises once they surpass this revenue threshold, similar to game engine licensing models.

While LTXV-13B is not yet capable of producing Hollywood-level movie quality, it holds immediate promise for animation and production workflows by automating resource-intensive tasks like key framing. This can significantly reduce production costs and accelerate creative processes.

Looking ahead, Lightricks envisions multimodal video models that integrate audio, music, and video in a shared latent space, simplifying complex tasks such as lip-syncing and enabling richer multimedia AI experiences.

LTXV-13B is currently available as an open-source release and is being integrated into Lightricks’ creative apps, including their flagship storytelling platform, LTX Studio, empowering creators with faster, high-quality AI video generation on accessible hardware.