Generative video is seemingly everywhere these days. Adobe has been a bit slow in the text-to-video rollout, but today is the day that Adobe’s “commercially safe” generative AI video model hits beta on the newly updated Adobe Firefly website. There is an impressive assortment of tools available now with more coming soon. Your existing Creative Cloud subscription gets you 1000 credits (more on that in a bit). I’m not that deep into the generative video craze (Jeff Foster is taking care of that here at PVC) but I do think Generative Extend in Adobe Premiere Pro can be a game changer and the most useful application of this technology so far.

I don’t have a ton of commentary, nor am I going to outline every detail in today’s big release, as you can read Adobe’s blog for that and dig deeper into all the different styles and things you can generate with Firefly. But here are the bullet points:

• Adobe Firefly Video Model, available in public beta, is the only generative AI video model that is IP-friendly and commercially safe, so brands and creative professionals can use it confidently for production-ready content
• New Firefly plan offerings unlock the full power of the new Firefly web app – the ultimate destination for generating images, videos, vectors and more with Firefly’s unmatched creative control
• Adobe Firefly provides the most comprehensive creative process with cross media workflows and direct integration with Adobe Creative Cloud apps to seamlessly move from ideation to production
• Leading brands and agencies including Dentsu, Gatorade and Stagwell are choosing Firefly as it provides the creative control needed to produce content at scale

I spent a little bit of time generating some video from random prompts as I sat in my home office on a rainy Tuesday evening. Generations are currently limited to around 5 seconds and take about 90 seconds to create.

There are a number of options you can use when generating video that helps with composition, shot size, and camera movement. Current outputs are limited to a 1080 H264, but that will most assuredly change as it comes out of beta. If Adobe expects us to use generated video for things like b-roll, then we’ve got to get different frame rate options, larger than 1080 media, and especially professional intermediate codecs like ProRes. I would expect all that to be available at some point when the Firefly generative video model fully ships. The model isn’t limited to just random shots as you can have it create things over a black background or a green screen, for example, to create some effect elements you could composite later. One option I’ve seen asked for is alpha channel support, which would make it really nice when creating those composite elements.

Firefly Video Model examples from a couple of generative video prompt engineering novices

Here are a few of the videos that I created. Once you create a video, there is a download button as well as your history. As far as I can tell, history is limited to the current browser session that you are in. That’s when I return later, my history was gone. I tried both Safari and Brave browsers, and I don’t think one worked any better than the other. It appeared to keep my generation history separate between those browsers.

When I was done, I was also curious about what a 13-year-old might create. So I sat my younger son down in front of the computer for a little while, gave him some very basic instructions on how to use Adobe Firefly generative video, and said, “Go for it.” You can see what a 13-year-old is interested in. If you want to see the full prompt, click over to YouTube where that’s in the description for most of the vidoes.

I went through a few generations trying to create a Firefly landscape. After a few attempts, I stopped at this one ⬇️

For this one I was inspired by looking out my window ⬇️

So those weren’t quite as realistic as I wanted them to be, so I thought I would try for something that could literally be used as a piece of B-roll ⬇️

Trying to create something a little bit funky, I wanted to use a reference photo for Image to video to make a piece of the cake come to life and move across the plate. While it didn’t move on its own as I wanted (click over to YouTube to see the full prompt) the result with the reference photo was quite impressive ⬇️

This is the reference photo I gave it ⬇️

As I mentioned, I let my 13-year-old sit down and have some fun with this as well. I gave him virtually no instruction on what to do. I just showed him where to type his prompt and told him to go for it.

Purple unicorns first ⬇️

And since he’s a basketball player, I knew basketball would make an appearance in his generations ⬇️

And this one is my personal favorite ⬇️

That was some fun stuff. When I was done, here’s what my Adobe Firefly would count as far as my credits.

That didn’t use very many of my credits, and I don’t know whether that’s because since it’s in beta, they’re not really counting the credits or if I’m grandfathered in due to some Adobe PR early access I’ve had to play with some of this stuff. But there are some new Firefly plans that are worth reading about as well. If you’re not used to seeing a “buy now” button on a service that you pay monthly for, well, get ready for that because AI generation isn’t free. The more you use it, the more you may have to pay.

I still want to know how Generative Extend in Adobe Premiere Pro will factor into these credits when it comes out of beta. There’s really a lot of fun stuff on the new Firefly website. In this day and age of questionably trained generative AI everywhere, I trust Adobe a bit more than a lot of the other companies out there regarding what they’re doing with generative AI. While the web tools make it easily accessible to almost anyone with the Creative Cloud sign-in, I’m looking forward to how some of these tools are going to integrate deeper into Premiere and After Effects. We’ll have to stay tuned for more on that.