Lightricks is separating its profitable Facetune business from its costly LTX AI video push, while cutting 75 more jobs. The move says plenty about the pressure now facing creative AI companies that grew fast before the business model became clear.
Lightricks is no longer trying to make one company carry two very different stories. The Jerusalem software maker behind Facetune is preparing to split its consumer app business from its LTX generative video division, a restructuring that comes with another 75 layoffs and a sharper question for the AI startup market: which part of the business can actually stand on its own?
According to Calcalist's CTech, the latest cuts affect 17% of Lightricks employees, including 55 workers in Israel, mainly in software and product roles. After this round and an earlier cut of 85 employees less than six months ago, the company is expected to have about 350 employees, roughly half the size it had in 2022.
That is a dramatic change for a company that once represented the creator economy boom at its most optimistic. Lightricks built a global consumer business through apps such as Facetune, Photoleap and Videoleap, then moved aggressively into generative AI video as the market shifted. In 2021, it raised a $130 million Series D at a $1.8 billion valuation, bringing total funding to $335 million.
The restructuring shows how much the ground has moved since then. Consumer creative tools are still valuable, but they now face competition from features bundled directly into phones, social platforms and general purpose AI products. Generative video, meanwhile, may be one of the most expensive categories in software, because every serious player needs models, infrastructure, researchers and distribution. That is a difficult mix for a mid-sized company to manage under one roof.
Facetune remains the engine that gives Lightricks room to maneuver. CTech reported that the app business generates around $300 million in revenue and is growing at about 20% a year. That is not a weak asset. It is a profitable consumer software business with a known brand, paying users and a clearer cost structure than frontier-style AI development.
But it is also a very different company from LTX. Facetune is about subscriptions, consumer retention, mobile product polish and brand familiarity. LTX is about selling AI video technology to enterprises, licensing models and competing with a field that includes OpenAI's Sora, Runway, Google Veo and a growing number of open-weight alternatives. Keeping both businesses together can make sense when one funds the other. It becomes harder when investors, employees and customers all need to understand what the company is really optimizing for.
Under the planned structure, co-founder and CEO Zeev Farbman will focus on LTX, while Asaf Porat, formerly deputy CEO of Outbrain, will lead the Facetune business. That leadership split matters because it gives each unit a more natural mandate. Facetune can be run for durability and cash generation. LTX can be run for enterprise adoption, model performance and future fundraising.
The layoffs also make the strategy more explicit. Farbman told Calcalist that Lightricks still has about 30 open roles, but that the AI modeling business requires different talent, specifically people who have built models before. That is the harder truth behind many AI restructurings. Companies are not simply shrinking. They are replacing one set of skills with another, often while asking the remaining business to finance the transition.
AI video is becoming an enterprise infrastructure bet
LTX began as a creative platform for AI video production, but the more interesting shift is now toward licensing. Lightricks has invested about $150 million into the AI business, and CTech reported that LTX is already generating annualized revenue in the tens of millions of dollars from model licensing, mainly from large technology companies and Fortune 500 clients.
That direction is practical. Running video models directly for customers can burn cash quickly, especially when usage scales faster than revenue quality. Licensing the technology to customers that can run models on their own infrastructure lowers Lightricks' compute burden and moves the business closer to enterprise software economics. It may be less flashy than consumer AI tools, but it is easier to defend if customers embed the models into real workflows.
LTX is also being aimed at physical AI, where video models can help train robots to understand human actions such as folding laundry. That is a more speculative market, but it points to why Lightricks does not want LTX judged only as another creator app. If video models become training infrastructure for robotics, simulation and enterprise media production, the upside looks very different from a subscription editing tool.
The challenge is that LTX is not alone. Open-weight video models keep improving, and the best-funded AI labs can spend heavily on research, infrastructure and distribution. Lightricks has used openness as part of its strategy with LTX model releases, but that cuts both ways: it can build developer interest while also making a durable product advantage harder to hold.
Lightricks is choosing focus because the market is forcing focus. If Facetune keeps producing cash and LTX proves that AI video can become a licensing business rather than a compute sink, the split could give both sides a better chance. If not, it will still serve as a warning to other creative AI startups: growth is no longer enough when the cost of staying competitive keeps rising.
Also read: Comfy Desktop makes ComfyUI easier for commercial AI builders • Anthropic's Mythos puts AI engineering economics under pressure • Bots now outnumber humans online and startups have to adapt
