Shutterstock pioneers the’research licence’ model with Lightricks and lowers barriers to AI training data.

Shutterstock has reshaped the way AI companies access training data by launching a novel approach called “research licensing” with AI creative technology company Lightricks. The partnership, announced yesterday, allows Lightricks’ open-source video creation model LTXV to be trained using Shutterstock’s extensive HD and 4-K video library.

This new licensing model addresses an important challenge in AI development – the high cost to access quality training data. It allows companies to begin with a small research license to test and experiment before committing to a more expensive commercial license.

Making ethical AI more accessible to startups

Many companies and model trainers are using unauthorized data scraping (19459042) to make the necessary investments to achieve the level of quality and trust needed to create commercially viable models, said Daniel Mandell in an exclusive interview VentureBeat. We don’t believe that financial investment should be an obstacle for those who want to enter this space ethically.

The two-phased approach could change how startups approach AI. Craig Andrews, Lightricks global PR manager, describes the move as “a turning-point for smaller, agile developers who want explore innovative applications of AI generative without the heavy upfront cost of traditional licensing.”

Legal protection and fair compensation at the age of AI

This timing is significant as it coincides with increasing legal scrutiny on AI training data practices. Several major AI firms are facing lawsuits for alleged unauthorized copyrighted use of material for model-training. Shutterstock offers a legitimate solution that ensures content creators are compensated.

Andrews explains that “we’re setting a new standard for ethical AI development, while ensuring creators are fairly paid for their work.” This approach not only fosters the creative ecosystem, but also establishes an sustainable framework for responsible AI innovations.

Revenue sharing: A win for creators and AI firms

Shutterstock implemented a revenue-sharing system where contributors get 20% of revenue from data licensing agreements. Mandell notes that only 1% of contributors have chosen not to use their content for AI training.

Lightricks intends to use the licensed data to enhance LTXV – its open-source video creation model released last month. According to Andrews, the model has already been a success. “Thousands of downloads have been made on GitHub and Hugging Face,” he said. One notable application is the generation of real-time videos for interactive ecommerce.

This partnership aims at addressing technical challenges in AI-generated video, especially motion consistency in longer videos. Andrews says that one of the biggest challenges in AI video creation is to achieve consistent motion and structure across longer video segments, without sacrificing quality. “Shutterstock’s high-quality videos library provides a large dataset that helps us to address this challenge.”

This partnership represents a significant shift in Shutterstock’s business model. The company has established partnerships with major AI firms such as Nvidia Meta and OpenAI. Mandell stresses that the research licensing model could democratize the access to high-quality data for smaller organisations and research institutions.

Setting new industry standards for ethical AI

This collaboration also reflects the growing trend towards transparency and ethical considerations when it comes to AI development. Lightricks opened-source LTXV to encourage collaboration and innovation while Shutterstock’s licensing approach ensures that content creators are properly compensated.

Mandell concludes, “The important message is that companies of all sizes and funding levels no longer have an excuse for scraping unlicensed content to use in training.” “There is a more effective way to enter this rapidly evolving market.”

The partnership could set a standard for how AI companies can access training data. This could influence industry practices, as concerns over the sources of AI data training continue to grow. This model’s success could determine if other content providers follow Shutterstock in creating more flexible and accessible licensing options for AI.