Oct 14 (Reuters) – Adobe (ADBE.O), opens new tab on Monday said it has started publicly distributing an AI model that can generate video from text prompts, joining the growing field of companies trying to upend film and television production using generative artificial intelligence.
The Firefly Video Model, as the technology is called, will compete with OpenAI’s Sora, which was introduced earlier this year, while TikTok owner ByteDance and Meta Platforms have also announced their video tools in recent months.
Facing much larger rivals, Adobe has staked its future on building models trained on data that it has rights to use, ensuring the output can be legally used in commercial work.
San Jose, California-based Adobe will start opening up the tool to people who have signed up for its waiting list but did not give a general release date.
While Adobe has not yet announced any customers using its video tools, it said on Monday that PepsiCo-owned (PEP.O), opens new tab Gatorade will use its image generation model for a site where customers can order custom-made bottles, and Mattel (MAT.O), opens new tab has been using Adobe tools to help design packaging for its Barbie line of dolls.
For its video tools, Adobe has aimed at making them practical for everyday use by video creators and editors, with a special focus on making the footage blend in with conventional footage, said Ely Greenfield, Adobe’s chief technology officer for digital media.
“We really focus on fine-grain control, teaching the model the concepts that video editors and videographers use – things like camera position, camera angle, camera motion,” Greenfield told Reuters in an interview.