Lightricks, renowned for its AI-driven innovations in visual content creation, has recently launched LTXV, a groundbreaking AI model capable of generating 60-second videos. This development marks a significant leap forward in the field of generative AI, offering creators unprecedented capabilities for long-form video creation. LTXV's unique architecture and features are poised to revolutionize various industries, from entertainment and advertising to education and beyond.
One of the key features of LTXV is its autoregressive streaming architecture, which enables continuous, real-time video generation. Unlike traditional AI models that are limited to short outputs, LTXV can generate videos exceeding 60 seconds, an eightfold increase compared to the previous industry standard. This allows developers and creators to build longer, coherent stories rather than just isolated clips. The model begins streaming results almost instantly, generating the first second of the video and then building the rest of the sequence on the fly. This real-time capability allows for interactive and continuous narratives with full control over scene development.
LTXV uses overlapping frame chunks to maintain continuity, ensuring that characters, motion, and storyline remain consistent throughout the video. This approach is similar to how large language models like ChatGPT generate text, but LTXV applies it visually, frame by frame. In a live demo, Lightricks showcased a continuous 60-second video featuring a woman cooking as a gorilla walks in and hugs her, demonstrating the model's ability to maintain narrative flow without abrupt transitions.
The model is not only powerful but also fast and efficient. LTXV reportedly delivers output much faster than competitors like Google's Veo 3, Runway Gen-4, or Kuaishou's Kling. It can be powered by a single Nvidia H100 or even high-end consumer GPUs. This is in contrast to other models that often require multiple GPUs to produce just a few seconds of high-resolution video.
Lightricks has made LTXV available as an open-source model on platforms like Hugging Face and GitHub. The model is free to use for individuals and small teams earning less than $10 million annually. This commitment to open development gives developers and independent artists the freedom to expand upon the core engine. The most advanced of Lightricks' models is LTXV-13B, which launched in May 2025 and introduced a novel multiscale rendering capability that enables creators to add more detail and color in a step-by-step fashion. The 13B model, which uses 13 billion parameters without compromising on processing speed, was trained using content licensed via partnerships with Getty and Shutterstock, for increased visual quality and ethics.
LTXV's capabilities open up new possibilities for various applications. It can be used to generate player-generated cutscenes, adaptive educational content, and real-time AR visuals synced with live performers. The model can also transform static images and text descriptions into dynamic videos. This feature is particularly beneficial for artists and designers looking to add a dynamic dimension to their creations.
Zeev Farbman, co-founder and CEO of Lightricks, believes that LTXV unlocks a new era for generative media. He emphasizes the model's ability to create longer scenes while maintaining full control of the extended sequences, enabling coherent storytelling with visual and semantic consistency. This transforms AI video from a demo or just a random clip into a true medium with creative intent.
Lightricks, founded in 2013, has a mission to bridge the gap between imagination and creation by bringing cutting-edge technology to the creative and business spaces. With LTXV, the company is not only pushing the boundaries of AI video generation but also empowering creators to bring their visions to life in new and exciting ways.
