Lightricks partners with Shutterstock to utilise its extensive video library for AI video model training.
- This collaboration introduces a unique ‘research licence’ reducing barriers for open-source model training.
- Lightricks aims to enhance its LTXV AI model using Shutterstock’s HD and 4K video resources.
- The partnership supports Lightricks’ mission to provide high-quality generative AI results.
- This move aligns with Lightricks’ strategy to make creative tools widely accessible through AI.
Lightricks has announced its collaboration with Shutterstock, leveraging Shutterstock’s expansive library of high-definition video content. This partnership is a groundbreaking initiative in AI training, allowing Lightricks to utilise Shutterstock’s resources to train its open-source LTXV video AI model. The use of high-quality HD and 4K videos is expected to enhance the model’s performance significantly.
The introduction of a ‘research licence’ by Shutterstock marks a novel approach that eases entry barriers for training open-source models. This licence option allows startups and AI companies like Lightricks to refine their AI tools with premium licensed data. It is a cost-effective strategy during the research and development phase, addressing a major challenge for many companies.
Zeev Farbman, co-founder and CEO of Lightricks, considers this long-term partnership with Shutterstock strategic, following previous licensing initiatives with major broadcasters. According to Farbman, this collaboration not only boosts their capability to train AI models with diversified and high-quality data but also enhances the relevance of their generative video tools for a growing community of professional creators and businesses.
Lightricks is committed to improving access to creative tools while ensuring high-quality outputs from generative AI, aligning this partnership with their ongoing mission. By integrating licensed video data from Shutterstock, Lightricks strengthens its training models. The focus is on robust datasets that respect creative professionals’ work while advancing open-source initiatives, as highlighted by the release of LTXV 0.9.
Daniel Mandell, Head of Data and AI Licensing at Shutterstock, expressed enthusiasm for the partnership. He remarked that the innovative research licence model allows partners to train AI models using licensed, reliable data right from the start, supporting long-term success. The vast library provided by Shutterstock, with over a billion diverse media assets, sets a high standard for multimodal model training.
This partnership underscores a significant advancement in AI training models, promoting broader access and superior quality in generative tools.
