Lightricks just made AI video generation 30x faster — and you won’t need a $10,000 GPU
Become a member of our daily and weekly newsletters for the latest updates and exclusive content about leading AI coverage. Leather
LighthouseThe company behind popular creative apps such as Facetune And Video capToday has announced the release of its most powerful AI video -oriented model so far. The LTX Video 13 billion parameter model (LTXV-13B) generates high-quality AI video up to 30 times faster than comparable models while they are performed on hardware of consumer quality instead of expensive Enterprise GPUs.
The model introduces ‘multi -display display’, a new technical approach that considerably increases efficiency by generating video in progressive layers of detail. This enables makers to produce professional quality AI videos on standard desktop computers and require high-quality laptops instead of specialized Enterprise equipment.
“The introduction of our 13B parameter LTX video model marks a crucial moment in AI-Video generation with the possibility of generating fast, high-quality videos about consumer gpu’s,” said Zeev Farbman, co-founder and CEO of Lightricks, in an exclusive interview with Venturebeeat. “Our users can now make content with more consistency, better quality and stricter control.”
How Lights AI video democratizes by solving the GPU memory problem
A major challenge for AI video generation has been the enormous calculation requirements. Leading models of companies such as Run” PikaAnd Luma Usually feed in the cloud on several GPUs with enterprise-grade with 80 GB or more VRAM (video memory), making the local implementation impractical for most users.
Farbman explained how LTXV-13B Addresses this restriction: “The most important dividing line between consumer and enterprise GPUs is the amount of Vram. Nvidia positions their gaming hardware with strict memory limits – the previous generation 3090 and 4090 GPUs maximum on 24 gigabytes of Vram, while the newest 5090 32 Gigaby, comparing, offers what a comparison.” Enterprise offers “
The new model is designed to work effectively within these consumer hardware restrictions. “The full model, without any quantization, without any approach, you can run on the top GPUs of the consumer – 3090, 4090, 5090, including their laptop versions,” Farbman noted.
Within ‘Multiscal Rendering’: the technique inspired by artist that makes AI Video Generation 30x faster
The core innovation behind LTXV-13BThe efficiency is the multi -rendering approach, which Farbman described as “the biggest technical breakthrough of this release.”
“It enables the model to gradually generate details,” he explained. “You start the coarse grid, get a rough approach to the scene, the movement of the moving objects, etc. And then the scene is a bit divided into tiles. And every tile is filled with gradually more details.”
This process reflects how artists approach complex scenes – starting with rough sketches before you add increasingly finer details. The advantage for AI is that “your peak amount of Vram is limited by a tile size, not by the final resolution,” Farbman said.
The model also has a more compressed latent space that requires less memory while retaining quality. “With videos you have a higher compression ratio with which, while you are in the latent space, you can just take less fram,” Farbman added.
Why Light Source Gambles on open source when AI markets are increasingly closed
Although many prominent AI models remain behind closed APIs, Lighttricks has made LTXV-13B fully open source, available on both Hug And Gitub. This decision comes during a period in which AI development of Open-Source has confronted challenges with commercial competition.
“Things were closed a year ago, but things open a bit. We see a lot of cool LLMs and diffusion models,” Farbman reflected. “I am now more optimistic than half a year ago.”
The open-source strategy also helps to accelerate research and improvement. “The most important rationality for open-sourcing is to lower the costs of your R&D,” Farbman explained. “There are a lot of people in the academic world who use the model, write papers and you start to become this curator who understands where the real gold is.”
How Getty and Shutterstock partnerships help to solve AI’s Copyright -challenges
As legal challenges register against AI companies using scraped training data, Lightrick’s partnerships has protected with Getty images And Shutter To gain access to License content for model training.
“Collecting data for training AI models is still a legal gray area,” Farbman acknowledged. “We have large customers in our business segment that give things like this, so we must ensure that we can offer them clean models.”
With these partnerships, Lightricks can offer a model with a reduced legal risk for commercial applications, which may give it an advantage in company markets that are concerned about copyright issues.
The Strategic Gamble: Why Lighttricks offers its advanced AI model for free for startups
In an unusual step for the AI industry, Lightricks LTXV-13B offers free license for companies with an annual income of less than $ 10 million. This approach is intended to build a community of developers and companies that can demonstrate the value of the model before income.
“Thinking was that the academic world is of the hook. These guys can do what they want with the model,” said Farbman. “With startups and industry you want to make win-win situations. I don’t think you can earn a lot of money from a community of artists playing with AI dings.”
For larger companies that are successful with the model, Lightricks is planning to negotiate license agreements that are comparable to how game engines charge successful developers. “As soon as they have reached ten million in income, we will talk to them about licenses,” Farbman explained.
Despite the progress represented by LTXV-13BFarbman acknowledges that AI video account still has restrictions. “If we are honest with ourselves and look at the top models, we are still far away from Hollywood films. They are not there yet,” he said.
However, he sees immediate practical applications in areas such as animation, where creative professionals can use AI to handle time -consuming aspects of production. “If you think about production costs of high-end animation, it is real creative work, people who think about important frames and the story, a small percentage of the budget. But Key Framing is a big thing,” Farbman noted.
Looks ahead, Farbman predicts that the next border is multimodal video models that integrate different media types into a shared latent space. “It will be music, audio, video, etc. And then things like doing good lip synchronization will be easier. All these things will disappear. You will have this multimodal model that knows how you can work in all these different modalities.”
LTXV-13B is now available as a open-source release and is integrated into the creative apps of Lighttricks, including the flagship storage platform, LTX Studio.
Source link
