How I created a 2 minute video using these 3 AI tools

Without the help of AI, frankly, I wouldn’t have bothered.

Today I’m going to show you how I chose and used three AI tools to improve short informational videos on my YouTube channel. One tool saved an entire small project, and the other greatly improved the end result. This project uses the latest versions of Adobe’s Photoshop Beta, Adobe Podcasts, and Apple. Final Cut Pro for Mac.

again: How does ChatGPT actually work?

project

i found a way to solve my problem glow forge, a laser cutter used in many maker projects. I am using it to make acrylic 2-color labels for 3D printed special adapters. The adapter attaches various tools to the dust collector I use in my workshop. This video was a short explanation of the problem (the resulting laser cut was the wrong size) and the solution (a checkbox in the options dialog).

In the video, I’m mostly sitting in front of the laser and talking, but there are pictures and screenshots sprinkled throughout to illustrate aspects of the solution.

Since this was a quick informational video, we didn’t put much effort into production design. To make matters worse, I ran into audio issues. Let’s start there.

AI #1: Clean Up Terrible Audio Using Adobe Podcasts

For some reason, my mic wasn’t recording clean audio when I was reading the script. I made this video mainly to help her find a solution in case the problem reappears and to share the solution with other Glowforge users of hers.

But it wasn’t worth going back and setting up the camera and prompter and doing the talking head bit again. Here’s an 8-second sample of the original audio:

As you can hear, it’s pretty rough. I’ve tried cleaning up other audio before, but the results were never good.

To get this working, I started with an original MP4 video recorded on an iPad. I used QuickTime to export the audio track (M4A file). Unfortunately Adobe Podcasts do not support M4A. I have a Creative Cloud license, so I went to Adobe Audition and opened the M4A and resaved it as WAV. (You can also use the free and open source Audacity program to do the same thing.)

again: These are my 5 favorite AI tools for work

Then I pointed my browser to Adobe Podcasts. This is Adobe’s experimental AI audio adjustment tool. (It’s not just for podcasters; I use it to fix the audio of YouTube videos.) This very useful little suite of text-to-speech and repair tools is free for everyone. can. Just sign up with your email address.

[音声を強化]Scroll down to Options,[試してみる]I pushed the button and uploaded the audio track.

Here’s what came back a few minutes later:

Wow, that’s an improvement! It’s the same day and night. All I had to do in my video editor was delete the original audio track and drop this one. The time it took was probably hours of frustration and not very good results, but I probably finished it in maybe 10 minutes overall. .

AI #2: Separate the performer from the background using Final Cut

When Final Cut Pro 10.6.6 was released on May 23rd, it included a new AI-powered masking tool, Scene Removal Mask. This tool provides green screen functionality without using green screen.

Here’s how it works: Capture 1-2 seconds of video with background only (no subject in frame). Then capture normal video. You can apply a scene removal mask to your video to remove the background behind the speakers. At this point, you can place anything behind the person in your video.

again: The Massive Adoption of Generative AI Tools Is Derailing One Very Important Factor

I started by creating a 2 minute video that looks like this:

Using the scene removal mask allowed me to remove the background. You can also zoom in on yourself to fill the frame better once the background is gone.

But there were still issues in the background that I wanted to fix. To that end, let’s switch to Photoshop’s new Generative Fill feature, then head back to Final Cut to wrap things up.

AI #3: Clean up the background with Adobe Generative Fill

This is what the camera caught behind where I was sitting and talking.

The first thing I wanted to do was remove the 3D printer blocking the corners of Glowforge.

Notice the printer in the image on the left.Draw a selection area around the printer and in Photoshop Beta[ジェネレーティブ フィル]clicked. Photoshop knew to remove his 3D printer and paint the missing areas of the wall. And here, to my surprise, the AI ​​drew the rest of his Glowforge that was originally missing from the image (including textures and lines).

You’ve decided the wall is a little empty, so why not use some old-school Photoshop features to add a poster? I added it to the background using features and blend modes.

again: 4 Best Free Video Editing Apps: Free But Powerful Software

We’ll return to AI Generative Fill in a few minutes, but first let’s synthesize the changes we’ve made so far.

Compositing video images

I went back to Final Cut and brought in the new background image that I modified in Photoshop. There are three track layers here. The lower layer compound clip integrates the previously repaired audio, the modified background image, and a masked version of the talking head scaled to fit the background.

But since the “masked” me was hovering over the background, I decided I could play around with the background. So, in the opening teaser clip, I applied a standard non-AI comic ink effect and placed myself on top of an ink-style background.

Then, at the end of the video, I replaced the background with my YouTube channel ID. Since we already had a masked clip of Talking Head, this was also very easy.

The video is now complete. But there was still one more place where AI needed help.

Move and fix blinds with Photoshop AI

When creating a YouTube video, the thumbnail of the video is often as important as the video itself, some say even more. Because it’s the thumbnail that grabs the viewer’s attention and gets them to watch the video.

The thumbnail format for my channel shows the letters on the left and bottom, and the image on the right. That image is usually a picture of me related to the content of the video. The problem was that the talking head was on the left side of the screen.

again: How to Use Photoshop: 5 Basic Steps for Editing Photos

This could have been easily fixed in Final Cut. But I wanted to use Photoshop’s new AI Background Remover feature instead. So I saved a frame from the video and brought it into Photoshop. I then masked the background using the new Photoshop AI background removal tool and slid the image to the right.

I then composited a background image (underneath my masked image, an image I had previously prepared for Final Cut) and added it to my channel’s branded style. It contained text about the video, a small robot logo, and a yellow sunburst bar along the bottom.

Unfortunately, Glowforge’s gray exhaust area (with mini Spock, Kirk, and Love Bear) got in the way in the thumbnail. YouTube thumbnails are much smaller than the image displayed above. The very small size did not look right. The red frame in the photo above shows the problem area.

All I really had to do was extend the white blinds down into the distraction area. Unfortunately, replicating the blinds was much more difficult than expected. I’ve tried using Photoshop’s clone stamp tool, cutting, pasting, and distorting parts of the blinds, but to no avail.

again: How to create Photoshop actions to share or sell

I think it’s because the blinds are not strictly parallel in the image. In reality they are parallel, but since the photo was taken at an angle, all the blind elements have slightly different shapes and angles.

So we went back to Photoshop AI’s Generative Fill feature. I turned off the text and my layers, selected the rectangle around the gray area, and hit generative fill. As you can see, the newly generated shades are almost perfect.

And here is the full video thumbnail. The clutter has been removed, a new shadow area added, and the image shifted to the right.

AI comes to the rescue

We are used to flashy special effects when watching big-budget blockbuster sci-fi movies and TV shows. But even the most mundane productions can have powerful special effects.

My video is only 2 minutes and 41 seconds long. It just shows how to select a checkbox in Adobe Illustrator. Yet, it uses a series of special effects designed to make the production look and sound even better.

But the problem is: This is such a generic video that there’s no justification for spending hours hiding the moving head and torso from frame to frame. In the background image he may have tried in vain to clone and duplicate the parts of the glow forge that are hidden by the 3D printer. I actually tried to extend the blinds using his traditional Photoshop tools, but gave up after about an hour.

again: Best AI Art Generator: Try DALL-E 2 and Other Fun Alternatives

And no matter how hard I try, I just don’t have the audio engineering talent to clean up audio that went very wrong during recording. But with the help of AI, we were able to clean it up in post-production.

Using each AI tool took 5-10 minutes. So, here’s an overview of how I used the AI ​​tools.

• Adobe Podcast Enhance Speech: Cleans up broken audio.
• Final Cut Pro Scene Deletion Mask: Removes the background, allowing the speaking head and torso to be scaled and moved (allowing for 3 custom backgrounds).
• Photoshop Generative Fill: Remove the ugly 3D printer and extend the background image glowforge.
• Photoshop Background Removal: Remove the head and torso from the background and move them to the right to make room for text.
• Photoshop Generative Fill: Enhance shadows to clean up messy images.

I used these in addition to my usual suite of tools in Final Cut and Photoshop that I’ve used for years. But knowing how difficult these five AI-assisted effects would have been without the AI, I probably wouldn’t have done them…and probably wouldn’t have published the video.

again: Final Cut Pro on iPad: an amazing and inspiring experience

Yes, I am concerned about how generative AI will affect employment (especially if many low-cost writing jobs are replaced by AI). But these 5 AI-generated fixes saved me tons of time, took care of tasks I didn’t have the skill set to do on my own, and created a more fun and engaging video. Because without these AI tools, I wouldn’t have been able to get it right.

Here’s how it all came together.

Have you ever used a generative AI tool to make a fix or tweak that you might not otherwise have been able to do (or justify your time working)? Let us know in the comments below.

---

Follow me on social media for updates on my daily projects. Please follow me on Twitter. @DavidGewirtzon Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, and on YouTube at YouTube.com/DavidGewirtzTV.