Adobe to pay $100+ to artists for their video clips, will be used to train Sora's competitor

Adobe is looking to expand its AI capabilities in video generation. According to documents seen by Bloomberg, the company is offering its network of photographers and artists $120 to submit short video clips of people engaged in everyday activities.

These clips are to be used as training data for Adobe’s AI systems. The types of videos requested include people walking, expressing emotions, interacting with objects like smartphones or exercise equipment, and even close-up shots of body parts like hands, feet, and eyes.

The compensation offered works out to roughly $2.62 per minute of usable video footage, though it can be as high as $7.25 per minute. This initiative comes amidst a growing trend of AI-powered creative tools. Adobe has already introduced features in Photoshop and Illustrator that allow users to generate images and illustrations based on text prompts. At least they would have something to prove, like they haven’t used videos in any unethical way, to which OpenAI’s CTO was tightlipped.

This has proven popular, with these tools being used billions of times so far. But, a OpenAI’s Sora has caused some to worry that Adobe could be left behind by this new technology. Also, Higgsfield.

While Adobe has confirmed they are working on video generation features with more details to be revealed later this year, the company’s current efforts clearly show the vast amount of data required to train AI models.

In contrast, Adobe has focused on training its AI models primarily on its extensive library of stock media content. When their stock library falls short, they have procured additional images directly from contributors. These contributions have previously involved large quantities of photos for specific subjects, with compensation ranging from 6 cents to 16 cents per image.

More here.