Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Join our daily and weekly newsletters for the latest updates and exclusive content in the industry’s leading AI coverage. Learn more
LilligoCompany behind popular creative applications Face and VideoPToday, the most powerful AI video generation model has been released today. LTX video 13 billion parametered model (LTXV-13B) GPUS creates a high-quality AI video, 30 times faster than comparable models, not expensive, not an expensive apparatus.
The model “Multispale Showing” presents a novel approach that increases the effectiveness of efficiency by creating a video with a detailed layer giving detailed information. This allows creators to produce professional quality AI videos in high-level laptops instead of requiring standard desktop computers and special enterprise equipment.
“13B parameter LTX application, at once in a sudden video generation, at a time, the exclusive interview with VentureBeat, which is one of the consumer GPUs,” users can now create content with more consistency, more consistency, more consistency, better. “
How to democratize the AI video by solving the GPU memory problem with Lighticks
A great problem for the generation of AI has been a great calculation requirements. Leading models from companies like Runway, Pikaand Luma Usually, 80GB or more Video Memory (video memory) is the impossibility of local placement for many enterprise levels at many enterprise levels, most ventures.
Farbman explained how LTXV-13B Touching on this limitation: “The main division line between the GPUS consumer and enterprise is the largest part of the VRAM.
The new model is designed to effectively operate this consumer hardware restrictions. “You will be able to celebrate the best consumer GPUS – 3090, 4090, 5090, 4090, 5090, 4090, 5090, 4090, 5090, 4090, 5090, 4090, 5090, 4090, 5090, 4090, 5090, without any approximation, without any approximation.
‘MultiCale Show’ Inside: AI video generation is a technique inspired by the artist who is 30x faster
Behind the main innovation LTXV-13BEfficiency is a multifacetment approach described as the “greatest technical advances in this issue”.
“This model allows you to gradually create details,” he said. “Starting from a rough network, a rough approximation of the stage, the movement of movements, etc.
This process begins the approach of artists to the complex scenes – gradually starting with rough sketches before adding delicate delicate details. Preference for AI “Vram’s summit is limited to a tile size, not the final resolution,” he said.
The model also has more compressed secret space that requires less memory while maintaining quality. “When you are in the secret space, you have a higher compression rate that allows you to get less vram,” Farbman added.
Why Lighticks bets on open source when AI markets are increasingly closed
Many leading AI models remains behind closed APIs, made Lighticks LTXV-13B Entirely open source, available in both Hug face and Entrusted. This decision comes in the period of open source AI development commercial problems.
“A year ago, things were closed, but the work opens a kind. We really see very cool LLMS and diffusion models,” he said. “I am more optimistic than half a year ago.”
Open source strategy also helps research and improve. “The main rationality for the open source is to reduce the value of R & D,” he said. “There are a ton of people in the academy that uses the model, write the documents and began to be this curator who understands where true gold is.”
How to help Getty and Shutterstock partnership to solve EU copyright problems
Lightricks provided the partnership, as legal problems against AI companies using valuable training information Getty pictures and Shutterstock to access the licensed content for model education.
“It is still a legitimate gray region,” said Farbman, “said Farbman,” said, “Data to collect information to prepare AI models.” “We have great customers in our segment, which cares for such items, so we need to make sure that we can provide clean models for them.”
These partnerships allow Lightricks to offer a model with reduced legal risks for commercial applications, potentially prefer the markets of copyright.
Strategic Gambling: Why Lighticks offers the advanced AI model for free
Lightricks in an unusual action for the EI industry, LTXV-13B is free for licenses to enterprises under annual income under $ 10 million. This approach aims to establish a union of developers and companies that can demonstrate the value of the model before monetization.
“Thought was the exempt of the academy from the hook. These children can do what they want with the model,” said Farbman. “You want to create a winning situation with the start and industry. I do not think that you can earn a ton of money from a society of artists playing with AI products.”
Lightricks for larger companies that succeed in the model plan to discuss license agreements in how successful developers of game engines. “Once you have received ten million income, we will come to licenses with them,” Farbman explained.
Despite the advances represented LTXV-13BFarbman AI acknowledges that video generation has restrictions. “If we are honest with ourselves and look at the top models, we are still away from Hollywood. It’s not there,” he said.
However, he sees practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical practical applications. “When you think about the production costs of high-level animation, people who think about the real creative work, key frames and stories are a small amount of budget. But the main framework is a large source.”
Farbman, who looked upon the next border, predicts that multimodal video models that connect different media to different media in a common secret space. “Music, audio, video, etc. will be easier as things will be easier like you will have and then do good lip synchronization. All these things will be disappeared. You will have this multimodal model that can work throughout these different modals.”
LTXV-13B is now available Open Source Issue and integrated into creative applications of Lighticks, including the flagship story platform, LTX Studio.
Source link
