How to Enhance Sets in Your Videos Using AI

0
81

[ad_1]

AI tools are here to help. Nearly every common piece of video editing software now has some degree of AI integration at this point though perhaps the most eye-catching results have come from Adobe’s Firefly beta.
Using generative AI we can now simple ask the software to make a change and it’ll use AI to create it almost seemingly out of nothing. It’s incredible.
One cool use case is to create more interesting backgrounds for your videos. Matti Haapoja has been using it to build virtual sets for his videos and the results are impressive. Learn how you can do it by watching this video.

Why might you want to build out some virtual sets? Well, unless you have a well-configured and organized studio space you are probably stuck with whatever empty room you have in your home or office. This can give you more options for how your background looks.
What Matti is doing is taking his normal A-roll footage and expanding the frame around him with more set dressing and detail. It looks surprisingly good and all he used was the new Generative Fill tools in the Adobe Photoshop Beta.
By taking a screenshot of his A roll to start he is able to bring that into Photoshop to start creating the new background.
Generative Fill is really cool since it allows you to make selections of the area and allow you to ask the software to edit it or fill it in with something completely new.
You’ll even get a few options to work with. For example, if you want a nice frame photo on the wall you can just select where, ask Photoshop, and then scroll through some options to find the best one.
Image Credit: Adobe
You can do more than just add a photo. Matti expanded the canvas and then started by expanding it quite a bit. He took care to get the ceiling looking good, added a window to one side, and then changed up the walls by adding decorations.
It’s actually cool since it was able to create a wide shot from a standard wide shot which gives him even more options for post.
Adding a variety of cropping by adding wider shots you can’t do easily with normal shooting techniques so this can help, especially if your actually space is tiny.
Another weird option is to shoot vertically and then use AI to fill in the sides for getting a standard widescreen image.
If you are creating thumbnails this software seems like a perfect option. It’ll help you get the framing just right and either clean up or add to your shot.
This is still the early days of AI, so there is a lot we haven’t even thought of for improving our video. It is getting to the point where they can start being used in actual workflows.
Unfortunately, being the early days it doesn’t always work well so Matti has some tips for getting it right.
The, hopefully, most obvious is that this is a lot easier with static shots. A roll on static backgrounds makes a lot of sense here.
The other thing to consider is how the image will look once you expand the frame. In his sample shot, Matti has a table which is a clean way to expand the frame and keeps his legs hidden so you don’t have more variables to work with.
Image Credit: Adobe
Consider the end result you may want. If you are looking for some sort of industrial vibe then if you find a concrete background it’ll give you a much better starting point for Adobe to work with.
Don’t go for everything all in one shot. It is much better to work with small pieces and build it up as you work. It has more things that can go wrong if you choose way too big of a frame.
Check out all the different options. It might help you think about other ideas or just provide a better version of what you asked for.
Take care to match up with your shot. In one scene he adds a window, but he made sure to add it only on the side where he has his light set up so that it looks like the light is coming from the window.
If you added a window to the wrong side the audience could pick up that it looks fake a lot more easily.
Have you tried out Generative Fill? What do you think about this virtual set technique?
[source: Matti Haapoja]
Order Links:

[ad_2]