Adobe is paying creators for videos to train its generative AI

According to a recent Bloomberg report, Adobe is paying creators for videos to help develop its generative AI. It seems that Adobe is keen to keep pace with OpenAI following that company’s recent demonstrations of its Sora app.

Adobe Firefly

Last year Adobe unveiled Firefly, the company’s first steps into the field of generative AI. This built on the Sensei AI that Adobe had been incorporating into its Creative Cloud apps for some time. The initial beta version of Firefly could only create high-quality images and text effects. However, at the start of this year, OpenAI announced its new text-to-video tool Sora. Soon after, OpenAI revealed that Sora will generate videos with audio. As a result, Adobe has some catching up to do.

Bloomberg report

Bloomberg reported that Adobe has been “offering its network of photographers and artists $120 to submit videos of people engaged in everyday actions.”  The content includes basic actions such as walking, as well as people interacting with objects “such as smartphones or fitness equipment.” In addition, Adobe wants videos showing people expressing emotions such as joy or anger, and simple anatomy shots of feet, hands, or eyes. Bloomberg said that Adobe wants more than 100 clips from creators. This means the $120 works out at between $2.62 and $7.25 per minute.

Why does Adobe want the videos?

Adobe is collating the videos to build up a library of material to train its generative AI. One of the main concerns with generative AI has been the source material used to train the software. If companies simply trawl the internet, then there is a risk that copyrighted material could be incorporated into the final work by the AI. However, Adobe says it trained Firefly on Adobe Stock images, openly licensed content and public domain content. Clearly, the company is pursuing a similar policy with the development of its text-to-video AI.

What we think

It’s reassuring to see that Adobe is continuing its responsible approach to developing generative AI. The text-to-video AI is being trained using videos that Adobe owns the rights to. Therefore, there shouldn’t be any concerns over copyright infringements with AI-generated content. However, the fact that Adobe is still asking for videos now does show that the company is lagging behind the competition a little. Adobe has said it will reveal more about its video-generating AI later this year. We will bring the news as soon as it is announced.

Pete Tomkies
Pete Tomkies
Pete Tomkies is a freelance cinematographer and camera operator from Manchester, UK. He also produces and directs short films as Duck66 Films. Pete's latest short Once Bitten... won 15 awards and was selected for 105 film festivals around the world.

Related Content