Adobe AI Firefly Video is Finally Here... Sort Of

Adobe Firefly Video Model
Credit: Adobe | Free use for news purposes

Adobe Firefly Video Model
Credit: Adobe | Free use for news purposes

Adobe has taken a significant step into Generative AI for video with the launch of its Firefly Video Model. This model was announced on October 14, 2024, at the world's largest creativity conference, Adobe Max.

Firefly is currently available in limited public beta and comes with several new features aimed towards streamlining the video editing process, mainly for professionals using Adobe's Premiere Pro and Firefly web app. While the excitement surrounding Adobe's debut into AI video is exciting, the initial offering has some flaws restricting its potential.

Generative Extend in Premiere Pro

Generative Extend in Premiere Pro
expand image
Credit: Adobe | Free use for news purposes

Generative Extend is one of the main tools integrated into Premier Pro in this beta release. It allows users to expand footage by up to two seconds, which is ideal for resolving minor issues like clumsy transitions or incomplete clips.

While it may not seem like much, two seconds could be quite beneficial when correcting shots that are too short, eliminating reshoots. It also helps correct mid-shot errors like minor movement issues or actors shifting eyeline.

Text-to-Video and Image-to-Video on the Web

The Adobe Firefly web app is where the real magic happens for creators eager to experiment with the text-to-video and image-to-video tools. These features empower users to generate short video clips with written prompts or still pictures, offering unprecedented control over the output through camera controls like angle, zoom, and motion.

The text-to-video feature allows the generating of a short video clip with a descriptive prompt. It functions similarly to other AI video generators like Runway and OpenAI's Sora. On the other hand, image-to-video allows animating still photos, producing short motion sequences that may be helpful when generating B-rolls or visualizing reshoots.

Early Adoption and Future Potential

Adobe's Firefly Video Model stands out for its commercial safety, which is a significant concern for professionals and brands concerned about copyright and licensing. However, Adobe guarantees its models are trained using licensed or public-domain content. This comes as a relief, as other AI video generators have faced scrutiny over their training data.

This makes the tool attractive to businesses needing to create large content without worrying about legal complications.

Early Adoption and Future Potential
expand image
Credit: Adobe | Free use for news purposes

Limitations of Adobe Firefly Video

While the Adobe Firefly Video Model marks an essential milestone in generative AI for video, the current beta version has several limitations. These make it difficult to use, particularly in professional-grade projects.

  • The clip duration is restricted to two seconds for Generative Extend in Premier Pro and five seconds for text-to-video and image-to-video on the Firefly app.
  • The resolution is limited to 1080p at 24 fps in Premier Pro and 720p for web tools.
  • The audio extension features can only reduce background noise or sound effects.
  • The Adobe Firefly generation time is about 90 seconds per clip, which could slow down workflows when working on large-scale projects.

While Adobe has promised future improvements, these limitations show that it will take some time before Firefly is ready for widespread professional use. As Adobe plans to refine these features and improve the Firefly video model, it could become a game changer for video creators in the future.


This Article's Topics

Explore new topics and discover content that's right for you!

NewsOpinionAI
Have an opinion on this article? We'd love to hear it!