When Prime Minister Benjamin Netanyahu speaks of Israel as a global leader in artificial intelligence, some in the local tech industry may scoff. Yet Lightricks, the Israeli AI company behind popular apps like Facetune and Videoleap, is proving that an Israeli firm can indeed compete among the world’s AI elite.
Lightricks on Thursday unveiled LTX-2, the latest version of its generative video model. The new system goes head-to-head with the newest tools from global giants; OpenAI’s Sora 2, Google’s Veo 3.1 and Chinese contenders such as Alibaba’s Wan and Kuaishou’s Kling.
A Hebrew-language video generated using Lightricks’ LTX-2 model
(Video: Lightricks)
'One-tenth the price of the big players'
“Just as Chinese developers built the DeepSeek LLM, a model with top-tier performance but at a local price point, we’re now doing the same for video, audio and image models,” Lightricks co-founder and CEO Dr. Zeev Farbman told ynet. “We’re presenting a world-class model that surpasses Sora 2 and Veo 3.1 in several aspects, at just one-tenth of their cost.”
According to company data, LTX-2 is the fastest video-generation model in the world. It can create 4K-quality video with fully synchronized sound from a simple text prompt, sometimes completing the process in less time than the duration of the video itself. The system’s operating costs are said to be about 40% lower than competitors’.
A Hebrew-language video generated using Alibaba’s Wan 2.5 model
(Video: Lightricks)
Lightricks, originally known for its photo and video editing apps for smartphones, shifted its focus in 2022 toward AI video technology. Its LTX Studio platform, based on the company’s proprietary models, allows creators to generate professional-quality video through text-based instructions.
In May, the company introduced LTXV, a previous version that used 13 billion parameters and featured a breakthrough compression technology, reducing video data size by a factor of 30, dramatically speeding up generation while cutting storage and processing costs.
A new era of AI-powered video creation
The LTX-2 engine represents Lightricks’ vision for transforming content creation. The model generates 4K video with original sound in seconds, synchronizing motion, dialogue and music into a cohesive output without separate editing. Using advanced AI rendering, it supports 48 frames per second and offers creators fine control over motion, depth and visual style.
LTX-2 doesn’t rely on costly cloud infrastructure, it can run locally on consumer-grade computers equipped with standard GPUs.
The model offers three performance modes: Fast – optimized for laptops, offering instant previews and live results. Pro – balancing speed and quality, aimed at content creators, marketers and studios. Ultra – delivering cinematic-quality results for premium productions.
A $600 billion market opportunity
The global video production market, estimated at around $600 billion, is one of LTX-2’s main targets. But Farbman says the model also has applications in gaming, film, advertising and animation.
He outlines a three-tier business model: licensing the model to large organizations for on-premise installation; offering API access for developers who want to integrate the technology into their own products; and providing consumer-level tools like LTX Studio for creatives.
“We’re building a full stack of tools because the true value of AI isn’t clear yet,” Farbman explained. “Eventually, every technology becomes a consumer product. The key is building products that compensate for the limitations of the tech itself, that’s where value is created.”
“The model alone won’t solve everything”
Farbman believes the industry is starting to understand that AI models alone aren’t enough. “At first, LLM developers claimed they would solve all problems. That didn’t happen. Now they’re focusing on code generation and AI search. The same will happen in video diffusion models like ours.”
While some in the industry are skeptical about whether AI-generated video can replace traditional film production, Farbman says the goal isn’t to replace cameras just yet. “Runway’s CEO once predicted that by 2025 we’d see a Hollywood film generated entirely with GenAI that would win an Oscar. We’re still far from that,” he said. “The hype got ahead of reality.”
Finding real-world applications
According to Farbman, the real opportunity lies in integrating GenAI into existing creative workflows. “Think about the animation pilot we’re developing with Disney,” he said. “Traditionally, creative teams would sketch key frames and send them overseas for the routine work. That’s the part GenAI can now handle.”
While OpenAI’s Sora may showcase “people riding dragons” for social media clips, Lightricks aims to embed its technology in production pipelines that already exist.
On Israel’s AI policy
Asked about government efforts to promote AI, Farbman said he’s seen encouraging signs. “A national committee was formed to explore AI policy, and they’ve spoken with me about it, which is a good start,” he said. “There are important questions, like whether academic researchers with limited budgets can get access to AI computing resources.”
However, Farbman said he doesn’t expect the government to fund AI infrastructure directly. “What they can do,” he added, “is bring investment from Gulf states and sovereign wealth funds that have already earmarked billions for AI. Help that money flow into Israel, and everyone benefits—economically, and maybe even politically.”