Those who work with generative video models know this well: there is a gulf between the promises of demos and real performance on a normal machine. The spectacular results shown on Twitter or in keynotes often only work in ultra-optimised environments, with $10,000 GPUs and pipelines impossible to replicate at home or in the studio. But now something has changed. Lightricks, already well-known in the world of mobile creativity, has released LTX-Video, an open-source model capable of generating real-time video on consumer hardware.
The news is sensational for those involved in digital content, NFT, metaverse or immersive experiences. LTX-Video produces video at 30 frames per second, in 1216×704 resolution, with a rendering speed exceeding the actual video duration. All without the need for data centres, expensive clouds or competitive NVIDIA infrastructures. Performance is also guaranteed on consumer GPUs, making this technology 30 times faster than previous standards. It’s a difference measured not in milliseconds, but in entirely new business models.
It is not just a technical issue. LTX-Video was trained on commercially licensed datasets, including Getty Images and Shutterstock. This means that the content generated is safer, cleaner, and above all legally usable by brands and creators. It is the first real step towards an AI video ecosystem designed for production and not just research or marketing.
The real strength, however, is accessibility. Until yesterday, creating AI video was a game for labs and funded startups. Today it could become a real tool in the hands of independent artists, content creators, web agencies and communication agencies, especially those already working in Web3. Think of what it could mean to be able to create AI-driven video experiences integrated in real time within virtual and dynamic spaces or immersive educational platforms, without depending on external vendors or prohibitively expensive licences.
For those who, like many Italian and European realities, have had to forgo AI video due to economic or infrastructural limitations, LTX-Video is a door that opens wide. But it is also a signal for the industry: the democratisation of generative creativity has begun, and it is much faster than we had anticipated. The giants selling AI as a service will have to rethink their plans, because if such a powerful model can run locally, the value shifts from the algorithm to the context of use.
The next viral content, the next AI short film, the next immersive interactive campaign will not come from those with the most budget, but from those with the most ideas and freedom to execute. And this time, no one will be able to hide behind the alibi of hardware anymore.
Here is the official source of the repository on GitHub with all the references to go deeper:
https://github.com/Lightricks/LTX-Video
