Adobe Acquires Video Clips to Develop AI Model

Adobe has commenced purchasing video clips to develop a “text-to-video” neural network, according to Bloomberg, citing sources.

Documents reveal that the tech company offers photographers and artists $120 for video clips depicting people engaged in everyday activities like walking or expressing emotions.

The primary aim is to train artificial intelligence. Previously, Adobe announced its work on video generation technology and plans to provide more details later this year, journalists noted.

Sources indicate that the company also requests simple anatomical images of feet, hands, or eyes. It seeks videos of people “interacting with objects” such as gadgets or exercise equipment. However, Adobe cautions against submitting copyrighted materials, nudity, or other “offensive content,” the media specifies.

Payment averages around $2.62 per minute of video.

Adobe aims to differentiate its models by primarily relying on its extensive library of stock media for marketers and creative agencies. When this proves insufficient, the company acquires images directly from creators.

The company has also offered payment for a large number of photos to train AI, such as images of bananas or flags. According to Bloomberg-reviewed listings, payments ranged from 6 to 16 cents per image.

Over the past year, Adobe has focused on integrating generative AI features into creative software. For instance, in May 2023, the company enhanced Photoshop with the Generative Fill neural network.

In March 2024, Adobe and Microsoft announced cross-platform AI integration to simplify the work of marketing departments.

Back in February 2024, Reddit secured a content licensing deal for AI model training.