Just two months after Lightricks announced its intention to split into two independent companies, the company has launched LTX, a new video model for its AI operations. The company now wants to present video not only as a final product for creating films, animation, or marketing content, but also as an infrastructure for training robots, operating avatars in real time, and building simulations.
The split in question is expected to be completed by the end of the year, subject to approval by the Israel Tax Authority, and will effectively separate Lightricks's profitable app activity, chiefly its Facetune photo editor, from LTX. Each company will have its own management, board of directors, and its own fundraising program. The move involved laying off 75 employees while creating plans for some 25 new AI-related positions.
LTX is now seeking to position itself not only as a developer of video-generation tools, but as a provider of a "world model" - a system that can assess a given situation, predict how it is likely to unfold, and help a machine determine its next move in the physical world.
"If you have a system that receives a frame and knows how to guess what the next frames will look like, it can also be used in robotics," Lightricks cofounder Dr. Zeev Farbman, who is now CEO of the new company, tells "Globes". "We want to build the expensive part, so that startups in this sector will have a foundation they can work with."
$50 million per model
"This year, there was a breakthrough following the publication of a paper called DreamZero," Farbman explains. "The researchers showed that a system that receives a frame and can predict what the subsequent frames will look like can also be used in robotics. A robot operating in the real world needs to predict what a particular action will look like and what its consequences will be, and that is the same capability underlying a video model." Unlike a video, however, a robot receives information not only from cameras, but also from touch, pressure, and temperature sensors.
According to Farbman, this data can be converted into a visual representation and fed into the model alongside the video. "The sensor readings are turned into an image in which each pixel represents, for example, the level of pressure or the temperature. We train the foundation model, and companies adapt it to their own data and applications."
Today, training a foundation model requires enormous amounts of data and computing power. According to Farbman, this is the part of the process that most startups cannot afford to finance themselves. He explains, "The most expensive part is training on all the data - huge amounts of video from the internet. We spend about $50 million a year on this. There are companies, particularly Chinese ones, that spend more."
"When you hear a startup say that it’s building a world model, you have to understand it doesn’t have a $50 million budget. As I said, we want to provide these companies with the expensive part: training the model. They can handle the adaptation to their own data and product themselves."
Farbman says one of LTX’s technological bets is on building a model that requires fewer tokens - the units of data processed by the model - thus reducing the amount of computing power needed to run it. "Smaller companies operate with more limited computing budgets; you need a model that a company can afford to run."
Driving widespread adoption
The fact that LTX allows users to download the model, run it on private infrastructure, and adapt it to their needs raises a central question: How does the company plan to make money from it? Farbman insists LTX does not want to rely solely on selling access to the model through an API, but instead plans to develop a licensing model that varies according to the industry and product. "As long as a company has less than $10 million in revenue, they don’t need to talk to us at all," he says. "They can use the model for any purpose. Once they exceed $10 million in revenue, they need to talk to us about licensing."
The licensing structure, he says, will not be uniform. In the case of an online video-generation service, LTX may receive a percentage of the revenue generated using the model. In the case of an animation company that runs the model on its own servers, the fee could be based on the production volume created.
This approach will initially allow LTX to expand the model’s use among startups and researchers without requiring them to pay upfront, and to begin charging only if their use of the technology develops into a business generating significant revenue.
At the same time, Farbman divides the model’s potential applications into three areas. The first, as noted, is media and entertainment, including the creation of animated videos and marketing content. The second is real-time video, such as virtual avatars that people can interact with in live conversations. "Instead of receiving an answer in text, you could have a conversation with a virtual lawyer who appears on video or even as a hologram. To do that, the model needs to generate video quickly and in real time," he explains.
The third area is, as noted, Physical AI, encompassing robotics and simulations and attracting interest from defense-tech companies as well. LTX does not intend to build robots itself, but plans to provide companies with the foundation model on which they can develop and train their own systems.
Working with everyone
In the world models sector, LTX is competing with companies with far greater resources, foremost among them Nvidia, which developed the Cosmos platform for Physical AI applications. Farbman argues, however, that the interests of hardware manufacturers themselves create an opportunity for LTX. "Most of the major players are choosing a closed approach; Nvidia is an exception." He adds that LTX wants to operate without being tied to one individual chip manufacturer.
"It’s very important to us that the model run on hardware from Apple, AMD, and Nvidia. That hardware flexibility is part of what differentiates us." At the same time, Farbman emphasizes the decision to release the model openly. "We want to build an open infrastructure everyone will want to use. We believe safety comes from many people being able to examine the technology to see what’s happening. Models that everyone can inspect are safer."
The new model is being launched, as noted, during a period of transition for Lightricks. Once the spin-off process is complete, Lightricks will comprise Facetune, its remaining mobile products, and influencer platform which connects creators with brands.
According to Farbman, there are currently fewer than 100 employees working at the mobile business, which generates more than $100 million in annual revenue. LTX, by contrast, has about 250 employees, most in R&D, and is continuing to hire staff for AI-focused areas. According to sources, once the split into two companies is complete, it will also be possible to seek out financial partners for the standalone business as well as potential new investors in LTX.
Published by Globes, Israel business news - en.globes.co.il - on August 19, 2026.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2026.