Adobe has started purchasing videos to create a neural network "text-video". This is reported by Bloomberg, citing sources.
According to the documents, the technology company is offering photographers and artists $120 to provide videos. They should depict people performing everyday activities like walking, or expressing emotions.
The main goal is to train artificial intelligence. Earlier, Adobe said that it is working on video generation technology and plans to talk about it in detail later this year, the journalists recalled.
According to sources, the company also asks to send simple anatomical pictures of legs, arms or eyes. She needs videos where people "interact with objects" like gadgets or exercise machines. However, Adobe warns against providing copyrighted materials, nudity or other "offensive content," the media said.
The payment averages about $2.62 dollars per minute video.
Adobe strives to differentiate its models, relying primarily on its own extensive library of stock media materials for marketers and creative agencies. In cases where this becomes insufficient, the company purchases images directly from the creators.
The company also offered payment for providing a large number of photos for AI training, such as images of bananas or flags. According to ads viewed by Bloomberg, the payment ranged from 6 to 16 cents per image.
Over the past year, Adobe has focused on adding generative artificial intelligence features to creative software. For example, in May 2023, the company added a Generative Fill neural network to Photoshop.
In March 2024, Adobe and Microsoft announced the cross-platform implementation of AI to simplify the work of marketing departments.
Recall that in February 2024, Reddit signed a deal to license content for training AI models.