Technology and software companies that are working on developing generative artificial intelligence platforms—in fact, those that aren't doing any development these days—are in dire need of content to help train their AI models. Adobe is now reportedly purchasing video to help develop its text-to-video conversion platform.
around bloombergAdobe is working hard to make a comeback in the generative video space, adding millions of 4K videos to its Adobe Stock library earlier this year and investing heavily in its Adobe Firefly generative AI platform. Similar to Firefly's text-to-image technology, Adobe wasn't the first to cross the finish line with video, and the company has only made general references to text-to-video so far. It's been several months since OpenAI's Sora was published, and many other companies are working on video generation.
Training an AI model to create still images is difficult, but training a video generation model is even more difficult. How OpenAI may have trained Sora has already sparked an outcry on YouTube, and OpenAI has been very closed off on the topic itself.
Adobe has a relatively ethical and open approach when it comes to Firefly, so it's no surprise that it would do the same with its rumored video platform. While some companies scrape content, Adobe is reportedly working on buying content from creators.
According to documents seen by Bloomberg, Adobe will “offer a network of photographers and artists $120 to post videos of people doing everyday activities like walking or expressing emotions like joy or anger.” ”. The goal is to procure assets for artificial intelligence training, the company wrote. ” bloomberg I will report it.
Based on the amount requested by Adobe and the length of the clip, the reported payment amount calculates to approximately $2.60 per minute for the submitted video, but could reach up to $7.25 per minute. There is a gender.
It's no surprise that Adobe is looking for specific types of content to help fill gaps in the Adobe Stock library. The Adobe Stock library is positioned as a commercial solution rather than a resource for all the different types of videos needed to train generative AI models.
Image credits: Header photo licensed via Depositphotos.
