Adobe has provided a preview of its upcoming Firefly Video Model that will be debuting later this year. The new AI-powered tool aims to streamline the creative process for video professionals, offering features that could significantly reduce production time and expand artistic possibilities.
The Firefly Video Model will introduce three key capabilities: Text to Video, Image to Video, and Generative Extend. These features are designed to help editors quickly ideate, fill gaps in their timelines, and add new elements to existing footage.
Text to Video allows users to generate short video clips from text prompts, with fine-grained controls over camera angles, motion, and zoom. This will be useful for creating B-roll or visualizing concepts before committing to a full shoot.
Image to Video can take still images or illustrations and transform them into live-action clips. Check out this example:
Perhaps most intriguing is Generative Extend, which will be integrated directly into Adobe Premiere Pro. This feature can add two seconds of AI-generated footage to the end of existing clips, potentially solving common editing challenges like smoothing transitions or extending shots to match audio cues.
Adobe distinguishes itself from other models by emphasizing that the Firefly Video Model is designed with commercial safety in mind. The company says that the model is trained only on content they have permission to use, licensed data and never on customer content.
While the Firefly Video Model may not be the most realistic AI video generator on the market, the early demos that Adobe shared are promising. Moreover the direct integration into professional tools like Premiere Pro could drive adoption.
The Firefly Video Model will be available in beta later this year, with Text to Video and Image to Video features accessible through Adobe's Firefly website, and Generative Extend integrated directly into Premiere Pro.