Adobe announced its entry into video generation with artificial intelligence by launching the new Firefly Video model during its “Adobe Max” conference.
Adobe offers a set of new tools that allow users to extend video clips and create new shots from still images and text descriptions through the video editing software “Premiere Pro.”
Among these tools, the “Generative Extend” tool stands out, available as a trial version in Premiere Pro, allowing users to extend the beginning or end of a short video clip, as well as make adjustments like correcting eye movement or dealing with unexpected motion. While this tool currently only extends clips up to two seconds, it can eliminate the need for reshooting to correct minor issues.
The tool allows for producing clips with resolutions of up to 720 or 1080 pixels at a speed of 24 frames per second, and can be used to enhance sound by extending audio effects and ambient sounds for up to ten seconds. However, the tool does not support extending spoken dialogue or music.
In addition, Adobe introduced two tools, “Text-to-Video” and “Image-to-Video” converters in a trial version. The “Text-to-Video” tool works by entering a text description of what the user wants to create, with the ability to select different styles such as realistic films and three-dimensional animation, and customize outputs using camera control tools to simulate different angles and movements.
The “Image-to-Video” tool allows inputting a reference image with a text description, providing more control over the results, and is useful for creating additional shots or envisioning reshoot scenes.
Currently, the generated clips using these tools can only be up to 5 seconds long, with a maximum quality of 720 pixels and a speed of 24 frames per second. For comparison, the OpenAI video generation model Sora can produce videos up to a minute in length but is not yet available to the public.
Adobe’s tools take about 90 seconds to produce a video, and the company is working on speeding up this process. Adobe mentioned that its tools are “commercially safe,” trained on licensed and authorized content, which may attract users considering other models’ scrutiny.
Finally, Adobe announced that videos created or modified using the Firefly model will be marked as using artificial intelligence while retaining ownership rights when published online.
