rare to see AI video This model is not technically equivalent to a closed black box, and it is even rarer to see a video model that is sufficiently optimized to run locally on the device without relying on cloud-based services. Lightricks-2, a new AI model video built in partnership with Nvidia, can do both.
Lightricks debuted this model in collaboration with Nvidia at CES 2026, one of the biggest technology trade shows. Nvidia also showed off a number of next-generation products powered by AI. Software updates for gamersan agent assistant, an education advisor, and an AI upscaler for smoother, sharper graphics.
Look at this: All announcements from the Nvidia Live CES 2026 stream
Lightricks' new model will be able to create AI clips of up to 20 seconds and 50 frames per second. This is on the longer end of the spectrum of AI video capabilities in the industry. This model also includes native audio. The ability to produce in 4K is important for creators who want to use new models for professional-level projects. But it's the features on the device that really separate the new model from competitors such as: Google Veo3 and OpenAI Sora.
This model was built with professional creators in mind, whether they are individual filmmakers or large studios. By focusing on clip quality along with on-device optimization, we aim to make clips one of the more attractive and secure options for creators interested in AI.
Learn more about CES hardware here. HP's new IT-friendly business laptops and AMD's fast mobile processors.
What is different about the new Lighttrix model?
When AI companies talk about an “open” model, they typically refer to: Zero gravity AI model. These models are not truly open source, so all parts of the process must be made public. However, developers can gain insight into how the model was built. Weights are like the ingredients in a cake. The open weight model will tell you all the materials used in the fabric, but not the exact measurements of each. Lightricks' model is open weight and is now available on HuggingFace and ComfyUI.
This is an example of the level of detail in LTX-2 AI video.
Lightricks new video models can also be run locally on the device. This is typically not the case with AI video. Generating even short AI video clips is a very computationally intensive process, so video models use more energy than other AI tools. For best results with most AI video generators, the data center computers used by Google or OpenAI should do the heavy lifting and generate the video in the cloud rather than on your laptop or phone. With Nvidia's RTX chips, you can get high-quality results without outsourcing your workloads to cloud services.
There are many benefits to running AI models locally. You control your data. There's no need to share it with big tech companies who might use it to improve their AI models. This is a very important factor for major entertainment studios. Dive into generative AI but Intellectual property needs to be protected right. Running your AI models on your device with the right equipment can yield faster results. On average, AI video prompts take 1-2 minutes to generate, so reducing this time can mean time and cost savings, two of the strongest arguments for creators incorporating AI into their work.
Click here for more information. AI memorization expands the AI wearable industry and New Gemini features coming to Google TV devices.
