Adobe is aiming to develop its own artificial intelligence model capable of converting texts into videos. Unlike other companies, they will offer a financial reward to creators whose content is used in training the model.
The use of smart technology in creating videos from texts seems to be the upcoming trend in the field of smart content production. OpenAI currently leads this field with its Sora AI model. However, Adobe differs by using video clips from various sources to train their upcoming smart model instead of relying on public sources like OpenAI, which may lead to legal issues due to copyright violations.
Adobe has requested ordinary video clips showcasing mundane activities like phone usage, people expressing emotions, and human body movements. Bloomberg reported that the company is purchasing up to 100 different types of video clips, with compensation ranging from $2.62 to $7.25 per minute, with each video valued at $120.
It is unclear who will be eligible to submit these clips to Adobe or if there is a limit on the quantity of clips that can be submitted. If one person can submit at least 100 videos at $120 each, it could lead to a decent compensation for capturing daily events.
In addition, participants must adhere to a number of rules. According to Bloomberg, the video must not be offensive, contain nudity, or violate copyright laws, which is standard practice for Adobe.
When the company ran a similar program to train its own AI image generator, images featuring trademarks, public figures, or even specific words were not accepted, as reported by Bloomberg. Similar rules could be applied to Adobe’s video generator training program.
However, copyright concerns likely pose a major challenge for Adobe. OpenAI has faced several lawsuits over ChatGPT, which creates summaries of copyrighted novels based on the training it reportedly underwent.
In an interview with the Wall Street Journal last month, OpenAI’s CEO, Mira Murati, stated that she does not even know if Sora has been trained on content from YouTube, which could lead to a violation of YouTube’s terms of service, according to the company’s CEO, Neil Mohan.
Legally, whether training an artificial intelligence model using copyrighted works is permitted remains undefined. Despite the cost of Adobe’s system, the company can obtain higher-quality images and fine-tune the upcoming video creation tool according to Adobe’s needs.
