News Summary:
Lightricks recently detailed its LTX-2.5, a 22-billion-parameter open-weights world model, which it released on August 11, 2026. This system generates synchronized video and audio from text, images, or existing clips, capable of producing a 10-second 720p video in 6.8 seconds on proprietary hardware. LTX-2.5 supports native 4K HDR and maintains character, lighting, and voice consistency across multiple shots. This development followed Lightricks' August 19 launch of LTX, its new video model for AI operations, just two months after the company announced plans to split into two independent entities. Lightricks now positions video as infrastructure for training robots, operating real-time avatars, and building simulations, beyond traditional media creation. On August 12, the company provided instructions for deploying LTX-2.5 on Runpod, recommending specific GPU hardware and noting gated access to its Hugging Face repository. Earlier, on August 8, Shani Mandel-Laufer, who joined Lightricks in 2022, was reported as a key contributor to building LTX, underscoring the company's shift from consumer photo and video editing apps to generative AI following the industry's technological revolution.
Subscribe for full access to Lightricks's profile
Subscribe for full access to Lightricks's products in full detail 
