Adobe is set to introduce a new generative AI-powered video creation and editing tool in a limited release later this year, the software company announced on Wednesday. The tool, named Adobe Firefly Video Model, will be released in beta and will complement Adobe's existing Firefly image-generating applications, which enable users to produce still images, designs, and vector graphics.
The new model will position Adobe in the expanding market for AI-based video generation tools, a sector already populated by competitors such as OpenAI's Sora, Stability AI's Stable Video Diffusion, and various AI video apps from smaller startups. According to Alexandru Costin, Adobe's vice president of generative AI, the tool can generate a five-second clip from a single prompt and can interpret both text and image prompts. Users can also dictate the desired camera angle, panning, motion, and zoom.
Costin emphasized that Adobe has invested heavily in ensuring the model meets the quality and prompt understanding expectations of videographers. "We've invested in making sure we really pay attention to the prompt ... respecting guidance from videographers much better than other (AI video) models," Costin told Reuters in an interview. Adobe clarified that the video model is trained on public domain or licensed content that it has permission to use, and not on any Adobe customer content.
Additionally, Adobe is launching Generative Extend, a tool that will be integrated into its Premiere Pro video editing software. This tool can extend any existing clip by two seconds by generating an appropriate insert to fill gaps in the footage. First previewed in April, the tool has received "a huge positive reaction from all of our customers," according to Costin.
