Adobe Shows Off Generative AI Video

Unless you have been living under a rock for the past year or two, you would be hard-pressed to have not seen something about AI on the internet. It's everywhere. AI has exploded onto the scene in just about every niche you can think of. 

When it comes to thoughts about AI the benefits and risks of AI, I have found that most people tend to fall into one of three categories.

Person #1 is vehemently opposed to it. Some because they are 100% convinced that this is how Skynet starts and it's only a matter of time before the Terminator shows up, and some because they have (far more legitimate) concerns about online security and intellectual property.

Person #2 thinks AI is the most amazing thing since the invention of pockets. People in this camp are willing to use it for everything from spitting out automation scripts and crunching numbers to generating music, images, and stories. 

Then there is Person #3. This person views AI in all its forms as a tool to enhance their existing workflow. AI doesn’t replace the need for a human, but if used correctly, it can make completing projects much easier.

As for me, Person #1 makes some compelling points (although I’m not so sure about the Terminator part), but I like to think I’m closer to Person #3. While we do need to find a way that AI is not stealing from human creators, AI is a tool that is made more powerful when used effectively and with a specific purpose rather than allowed to run free-range.

I bring all this up because just this week, Adobe released plans to add AI features, including generative AI, to Adobe Premiere Pro.

Most of these features deal with video editing rather than photography editing, but because I’ve been dipping my toes more and more into video editing lately, I wanted to share my thoughts on these features and what they could mean for the future of AI video.

Generative Extend 

Adobe describes the Generative Extend features as a way to “seamlessly add frames to make clips longer, so it’s easier to perfectly time edits and add smooth transitions.” 

One thing I quickly learned when I started video editing is how important it is to switch between clips with perfect timing. When you’re out collecting raw footage, it can be hard to know exactly how long you will need to stay on a certain shot in the final video. From the sounds of it, Generative Extend should make it a lot easier to fine-tune that timing, even if your clips are a hair too short. 

I’m excited to play with this feature and get a feel for how smooth AI-extended clips are or if it’s a little obvious they’ve been edited.

Object Addition & Removal

This is a popular feature these days. As the name implies, it allows a video editor to completely remove unwanted objects from a shot and add new objects. I find this kind of control over what is in-frame really intriguing. 

Text to Video

This is the one I am most excited about. When you need a little B-roll, you type in a command and Adobe generates a clip you can easily add to your video. 

I tend to underestimate just how much B-roll I need for a video and though I’m getting better, it will be nice to know I have options waiting for me in the editing room, so to speak. That said, we will see how good the AI-generated Broll turns out. A lot of the Broll I do use is of people goofing off together or throwing themselves down the side of a mountain on a bike – I’m not so convinced AI can reliably generate clips of that kind, but I’m willing to play around with it.

While I don't expect these features to be perfect out of the gate, after seeing what Adobe managed to do with features like these in Photoshop, to tell you I'm excited would be an understatement. Being able to generate b-roll footage directly in Premiere with AI or use that AI to enhance my audio for me auto-magically – sign me up!

With both photos and video, having a fast and efficient workflow is critical for me. Features that promise to speed that up are a pretty reliable way to get me excited. 

I'm also happy to see that Adobe has decided to work with third-party AI models instead of relying on just their own. This gives the users the ability to work with models they already know as well as access to ones that may be more specialized to their needs.

The new features are set to be released later this year and you can bet I will be back to talk about them – as soon as I’m done playing with them.