Adobe has offered a sneak peek into its upcoming generative AI tools for video, including a new feature that allows for the creation of video clips from still images. This comes as part of the evolving Firefly video model, which was first showcased in April. Firefly is expected to power a range of AI-driven video and audio editing capabilities within Adobe’s Creative Cloud suite.

In its latest teaser, Adobe revealed the potential of Firefly’s text-to-video feature, which was announced earlier this year, though not yet demonstrated. This tool enables users to create video clips based on text descriptions and further refine the output using various “camera controls,” allowing adjustments to angles, movement, and shooting distance. Additionally, an image-to-video capability was highlighted, where video clips can be generated from reference images, a tool that could prove handy for filling production gaps or creating supplementary B-roll footage.

If the showcased footage is anything to go by, the quality of these generated videos is comparable to what OpenAI’s Sora model has delivered so far. Notably, Adobe is considering integrating Sora into its Premiere Pro video editing software as a third-party tool. However, a current limitation is the duration of the videos produced by Firefly’s text-to-video and image-to-video features, with a maximum length of just five seconds. Adobe’s VP of generative AI, Alexandru Costin, confirmed this detail to The Verge.

One significant advantage Adobe’s Firefly model might hold over competitors like Sora is its focus on being “commercially safe.” Firefly has been trained using openly licensed, public domain content, as well as Adobe Stock materials, potentially mitigating concerns around copyright infringement—an issue that has plagued other AI models.

Both the text-to-video and image-to-video features are set to launch in beta as part of a standalone Firefly app later this year. Over time, Adobe plans to integrate Firefly into its Creative Cloud, Experience Cloud, and Adobe Express platforms.

In addition to these tools, Adobe also teased a new “Generative Extend” feature for Premiere Pro, which can lengthen existing video footage in a manner similar to Photoshop’s Generative Expand tool that extends image backgrounds. This new feature is also expected to roll out at some point later this year.