It is uncommon to see an AI video mannequin that is not the technological equal of a closed black field, and it is even rarer to see a video mannequin that is optimized sufficient to run regionally in your units somewhat than counting on cloud-based companies. Lightricks-2, a brand new AI mannequin video inbuilt partnership with Nvidia, can do each.
Lightricks debuted the mannequin with Nvidia at CES 2026, one of many greatest tech commerce reveals. Nvidia additionally confirmed off a lot of AI-powered and next-generation software program updates for players, together with an agentic assistant, an academic advisor and an AI upscaler for smoother, extra outlined graphics.
Watch this: Each Announcement from the Nvidia Dwell CES 2026 Stream
Lightricks’ new mannequin will be capable to create AI clips which might be as much as 20 seconds lengthy, 50 frames per second — that is on the longer finish of the spectrum of the business’s AI video capabilities. The mannequin can even embody native audio. The power to generate in 4K shall be important for creators who wish to use the brand new mannequin for professional-grade initiatives. Nevertheless it’s the on-device capabilities that actually set the brand new mannequin aside from rivals like Google’s Veo 3 and OpenAI’s Sora.
The mannequin was constructed with skilled creators in thoughts, whether or not that is particular person filmmakers or huge studios. The deal with the standard of the clip, together with its on-device optimizations, goals to make it one of many extra interesting and safe choices for AI-inclined creators.
For extra on {hardware} from CES, take a look at HP’s new IT-friendly enterprise laptops and AMD’s speedy cellular processors.
What’s completely different about Lightricks’ new mannequin
When AI corporations discuss “open” fashions, they’re usually referring to open-weight AI fashions. These fashions aren’t really open-source, which requires each a part of the method to be disclosed. However they do give builders insights into how the mannequin was constructed. The weights are like substances in a cake; open-weights fashions let you know all of the substances that went into the batter, however you do not know the precise measurements of every. Lightricks’ mannequin is open-weight and out there now on HuggingFace and ComfyUI.
That is an instance of the extent of element in LTX-2 AI movies.
Lightricks’ new video mannequin can also be capable of run regionally in your units. This isn’t usually the case for AI video. Producing even brief AI video clips is a really compute-intensive course of, which is why video fashions use extra power than different AI instruments. To get one of the best outcomes with most AI video mills, you are going to need knowledge middle computer systems utilized by Google or OpenAI to do the heavy lifting and generate your movies within the cloud somewhat than in your laptop computer or telephone. With Nvidia’s RTX chips, you may get these high-quality outcomes with out outsourcing the workload to a cloud service.
There are a whole lot of advantages to working AI fashions regionally. You are answerable for your knowledge; you do not have to share it with huge tech corporations which will use it to enhance their very own AI fashions. That is a particularly essential issue for giant leisure studios which might be diving into generative AI however want to guard their mental property rights. Working AI fashions in your gadget, with the right tools, may also offer you outcomes quicker. The common AI video immediate takes 1-2 minutes to generate, so decreasing that would assist save money and time, that are two of the strongest arguments for creators integrating AI into their work.
For extra, take a look at the AI note-taking ring increasing the AI wearable business and the new Gemini options coming to Google TV units.
