ChatGPT Codex Turns Raw Clips Into a Polished Video (Full Transcript)

[00:00:00] Speaker 1: Today, we'll learn how to edit videos with ChatGPT Codecs. Here, I have a folder with 16 unedited video clips that I filmed showing how to make chocolate chip cookies. Altogether, it's about three and a half minutes of raw footage. Normally, I bring all of this into software like DaVinci Resolve, cut the clips together, add text overlays, captions, and more. But instead, let's give all of this to ChatGPT Codecs to see if it can turn it into a polished video that's ready to post. And a quick disclosure, OpenAI is sponsoring this video, but as always, I'll show you exactly how it works and the results that I get. And I'm not going to tell it where to make every cut or how to put everything together. I want to see how much of the editing it can actually handle on its own. I'm Kevin, and let's dive in. To get started, head to the following website. You'll find a link right here at the bottom of the screen. You can download the app for both Windows and Mac. Once you install the app and sign in with your ChatGPT account, over on the left-hand side, select Codecs. Now, you might be wondering, why are we using Codecs instead of just a regular ChatGPT conversation? I'm using Codecs because we're working with an entire folder of local video files, and we wanted to actually build, render, test, and revise the video project. This is exactly the type of longer, multi-step task that Codecs is designed for. Now, you've probably noticed Codecs doesn't have a traditional video editing timeline like you'd find in DaVinci Resolve or Adobe Premiere Pro. So, to help it create our video, we're going to add a plugin called HyperFrames. Plugins can extend what Codecs can do by giving it additional tools and specialized instructions for different types of work. Over in Codecs, let's open Plugins, and right up here, let's search for HyperFrames. There it is. Let's install it. HyperFrames gives Codecs the tools and instructions it needs to work with video, audio, captions, graphics, and animations, and then render everything into a finished video. All right. Now that we have HyperFrames installed, let's create a new project. Over on the left-hand side, let's hover over Projects, and right over here, you'll see the plus icon. Let's click on that to create a new project. Now, right up here, I could enter in a project name. I'll call this Chocolate Chip Cookies, and right down below, I can now point this to a source folder on my computer. Let's click on Add, and right up here, I'll select this folder, which contains all of my raw video files. As a note, if you'd like to follow along, I've included a link to all of these files in the description down below. And over here, I'll click on Select Folder. Now, before we get started, let me show you a few of the original clips so you can see what we're working with.

[00:03:05] Speaker 2: Here's how you can bake the best chocolate chip cookies in just 60 seconds.

[00:03:10] Speaker 1: As you can see, these are completely unedited. There are pauses, extra footage, and nothing has been put together yet. I'm also intentionally not organizing or labeling any of these clips. I want to see if Codex can look through the footage, understand what's happening, and figure out how to put everything together. Right down here, let's click on Create Project. Right up here, let's click on Trust Folder. Over on the left-hand side, let's now expand Projects, and there's a new project called Chocolate Chip Cookies. Over here, let's kick off a new chat. Right up top, let's minimize the sidebar. Now comes the fun part. Let's see what Codex can do with all of this raw footage. I'm going to keep my prompt relatively high level. I don't want to tell it which clips to use, where to make cuts, or exactly where to put everything together. I want Codex to make those editing decisions on its own. Right down below, I'll paste in my prompt. Feel free to pause it if you'd like to read it, and I've also included it in the description down below if you want to try this exact same prompt. Now, you'll notice here at the very bottom of the prompt, I refer to HyperFrames. That's the plugin that we're using. I'm going to at mention it here. I'll enter in the at character, and there we see HyperFrames by Haygen. I'll select this, and now I've inserted that into the prompt. This will make sure that Codex uses this plugin to pull together the video. Now that I've entered in the prompt, I've also referred to the plugin that I want to use. Right over here, let's now send that and see what it comes up with. Now, this is a pretty substantial task, so Codex is going to need a little bit of time. It's analyzing the footage, figuring out what to keep, and how to sequence everything, and then it's going to use HyperFrames to build and also render the finished video. I'll let it keep working, and then we can come back when the first version is ready. All right, Codex has finished rendering the video, so let's take a look at what it created. Now, remember, all I gave it was a folder with 16 raw, unlabeled video clips, and a description of the finished video I wanted. Let's watch.

[00:05:11] Speaker 2: Here's how you can bake the best chocolate chip cookies in just 60 seconds.

[00:05:40] Speaker 1: For a first pass, that's really impressive. It figured out the story. It cut everything down from about three and a half minutes to under a minute. It added all the ingredient callouts. It also added captions, and it added some music as well. And it did all of that from the original raw footage and a prompt. This is already a really good first edit, but I think we can make the opening even stronger. So right here, I'd like to include some type of thumbnail or graphic. And right down here, instead of typing another prompt, we can use voice to give Codex feedback while I'm working. So over here, let's start voice chat. Let's use ImageGen to create a thumbnail-style opening graphic, and why don't we feature the finished chocolate chip cookie in that, and maybe we throw in some text that says, the best chocolate chip cookies. Let's use hyperframes again to add it briefly to the beginning of the existing video, so maybe the first frame or two. And let's not change anything else.

[00:06:44] Speaker 3: Okay, let me check that. I'll use the finished cookie frame as a reference, and the graphic will only cover the first two video frames, so the existing timing and audio stay exactly the same.

[00:06:55] Speaker 1: Perfect, that sounds good.

[00:06:56] Speaker 3: The generated graphic has the finished cookies and the exact title, and it's limited to the first two frames. Everything else is unchanged. I'm just checking the render, then I'll share the review video. The two-frame intro passed validation, and the updated video is rendering now with the previous export still available. I'll send the new review link as soon as it's ready. The updated review video is ready. The title card shows for only the first two frames, and everything else, including the audio, is unchanged.

[00:07:25] Speaker 1: All right, Codex finished the changes, and over on the right-hand side, we can see the updated thumbnail for this video, and that looks pretty good, and it matches the text that I specified. Looks great. And right down below, if I click on play, here's how you can bake. There, it jumps into the rest of the video that we already had. Now, of course, I could keep giving Codex feedback and continue refining the video, but for now, I'd say this is pretty darn good. And that's how you can use Chatsubt Codex to edit a video from start to finish. Overall, I'm impressed. It understood what was happening in each clip. It figured out how to sequence everything. It added relevant text and captions, brought in music, and it even created an opening graphic. Now, of course, there's still plenty that I could refine, and I'd always review a video before publishing, but this could save a lot of time when working on creative projects. Of course, Codex can do a lot more than just editing videos. If you want to try it yourself, I've included a link in the description down below. Thanks for watching, and I'll see you in the next one.