Shutterstock Now Offers Research Licenses For AI Video Model Training

Shutterstock has unveiled an ethical licensing model designed to reshape the way AI developers access high-quality training data.

The new Shutterstock research license allows AI startups and smaller developers to train AI systems affordably and responsibly. Its first application is powering LTXV (LTX Video), a real-time video generation model developed by Lightricks.

By utilizing Shutterstock’s vast library of HD and 4K videos, Lightricks aims to overcome technical barriers in AI video production while adhering to ethical standards. With a vast library of over 800 million images and 55 million video clips, Shutterstock currently serves approximately 4.1 million active customers

“Many companies and model trainers have taken the route of unauthorized data scraping [instead of] making the necessary investment to achieve the quality and level of trust needed to develop commercially viable models,” said Daniel Mandell, Shutterstock’s global head of data licensing and AI.

“However, we don’t think that financial investment should be a barrier for those looking to enter this space with an ethical approach.”

Shutterstock’s Research License: A Two-Phase Licensing Model

Shutterstock’s research license introduces a two-phase licensing framework aimed at addressing the prohibitive costs associated with AI training. In the initial phase, developers can experiment with smaller datasets for testing and validation.

Once their projects are ready for commercialization, they can transition to full commercial licenses. This flexible approach is designed to lower the financial barriers to entry for startups and smaller organizations.

Mandell emphasized that Shutterstock aimed to establish ethical standards in AI development and ensure fair compensation for content creators.

One of the most persistent challenges in AI video generation is maintaining smooth, coherent motion across frames, especially in longer clips. LTXV’s architecture is specifically designed to address this issue. By training on Shutterstock’s diverse and high-quality video datasets, LTXV achieves a level of motion consistency that sets it apart from competing models.

“Shutterstock’s data enables us to train models responsibly and ethically,” said Zeev Farbman, CEO of Lightricks. “This partnership ensures we respect creative rights while driving forward the technical capabilities of our tools.”

Fair Compensation for Contributors

Contributors to Shutterstock’s platform benefit from a revenue-sharing model, receiving 20% of the revenue generated from data licensing agreements. Additionally, content creators can opt out of having their work used for AI training purposes, a choice exercised by less than 1% of Shutterstock contributors.

Mandell attributes this low opt-out rate to the transparency and fairness of the licensing model.

The initiative positions Shutterstock as a driver of promoting ethical AI practices at a time when the industry faces growing scrutiny over data sourcing. Multiple lawsuits have been filed against AI companies accused of using copyrighted material in their training datasets without authorization.

Capabilities of LTXV: Real-Time AI Video Generation

LTXV, developed by Lightricks, can produce videos faster than playback speed, generating 24 frames per second and delivering high-resolution videos in as little as 30 seconds on high-end hardware.

LTXV offers multiple modes of operation, including text-to-video, image-to-video, and video-to-video generation. These features allow creators to generate animations, transform existing footage, or create entirely new videos based on descriptive prompts. The flexibility of these tools makes LTXV accessible to both professional filmmakers and independent creators.

Technical Innovations

The model is built on a Diffusion Transformer (DiT) architecture with two billion parameters, which enables it to achieve smooth transitions between frames and maintain consistency over extended clips.

LTXV’s default output resolution is 768×512, but it can scale up to 720×1280 while retaining performance. Unlike many AI models, which require powerful infrastructure, LTXV can operate efficiently on consumer-grade GPUs with as little as 6GB of VRAM, broadening its accessibility to smaller studios and individual developers.

Craig Andrews, global PR manager at Lightricks, stated that motion consistency was one of the biggest technical challenges in AI video generation and explained that Shutterstock’s video library played an essential role in addressing this issue.

LTXV is released under the Apache 2.0 license, making it freely available for modification, use, and redistribution. This open-source framework encourages collaboration among developers and researchers while ensuring accessibility for independent creators and small businesses. The model is hosted on GitHub and Hugging Face, where it has already been downloaded thousands of times.

In addition to its open-source availability, LTXV integrates with platforms like ComfyUI, enabling users to customize and adapt its features to specific needs. This level of flexibility makes LTXV a valuable tool for both technical experts and creative professionals.

LTXV’s ability to generate extended clips with consistent quality makes it highly scalable for various applications. From interactive e-commerce and advertising to game development and content creation, the model’s versatility opens new possibilities for creators. Its compatibility with consumer-grade GPUs further broadens its appeal, ensuring that even small studios can harness its capabilities.

Shutterstock’s licensing framework provides a compliant alternative that benefits both developers and content creators. By leveraging Shutterstock’s ethically sourced datasets, LTXV exemplifies how open-source innovation can align with ethical practices.