A creator uses Descript AI Agent to edit an entire video by typing commands in plain English. The tool removes filler words, adds titles, cleans audio, and inserts B-roll automatically, showing what AI editing can do in real time.
AI made this for me. I didn't do that. It automatically put in that title. Slide it put in that one. Now you can edit a YouTube video just by typing what you want the software to do. In this video, I'll show you how to use Descript AI Agent to edit faster than ever. Inside my brand new project, I'm just going to find the video that I'm going to use, my RAW video. I'm going to drag that in. You can see it's a 15 minute and 18 second video and there's a lot of stuff that needs to be cut. Like right here I was saying test, test, test and I accidentally left the recorder going. So there's a lot of stuff that's going to need to get cut out here. So to edit this, I'm going to first come over to where it says AI Agent. And by the way, this is in beta, so you won't see this unless you've applied and been accepted to test out this feature, which you can do using the link in the description. So at the bottom there's a text box where I can start to give instructions to the agent. So I'm going to say edit this video for content, remove gaps, filler words and retakes. And then I'm going to click submit. It says thinking. It tells me it's planned so how it's going to attack this instruction. Okay, and it's done. It tells me it completed that in 34.4 seconds. It tells me what it did. All gaps have been shortened or removed. Filler words, removed, retakes where the speaker restarted or corrected themselves have been cut. Pre and post recording. Chatter has also been removed. And so you can see over on the transcript all this stuff that's lined out. And now it's just gonna start on my first good take. So everything else before that, which was like six minutes of content is gone. Awesome. And then like right here, the chances are you found me from the D Scriptmaster YouTube channel. And then it crossed out that first time I said it, which is a retake. So it took out the bad take, correctly identified that. Very cool. Next I'm gonna tell it add text titles to this video where it makes sense and would add to the clarity of the message. Submit. Okay, and it's done. It says it took 133 seconds. And notice over on the left side, it automatically created a bunch of scenes throughout the video. And if I click on some of these scenes, there's our opener. And again, AI made this for me. I didn't do that. So it automatically put in that title slide. It put in that one. It sprinkled in a bunch of these lower third titles throughout the video. It put in a couple left half slides and in total, it split up the video into 20 different scenes. This very cool. Now I'm going to say clean up the audio on this entire video submit. So it came up with a plan. So go to the layer panel in your project. Select Studio Sound. Studio Sound will automatically remove. So this is interesting. Rather than actually doing it, it's just telling me how to apply Studio Sound. Okay, that's fine. Let's go to the next one. I'm going to say add B roll to supplement this video and make it more interesting. You can see it's starting to put in scenes with placeholder layers on the left side. And then in a moment, it's going to replace that with actual B roll. And check that out. That was added by AI. We are entering an era where all you need to edit videos is be able to type in plain English what the AI needs to do. Okay. And it's just finished up its work. That time it took 512 seconds, so quite a bit of time. And I can see it's not perfect. There's a couple scenes where the placeholder layer is still in there, but that's no problem. I could click on it, say add media and then put in my own stuff. But overall, pretty incredible that it was able to do that. And with all of this, of course, you as the human should go through and check everything because this whole thing is in beta. But you should check it anyways and make sure that everything is flowing smoothly and that it's up to your normal standard of edits. And with that said, above every prompt in the chat history, there's a button that says roll back to checkpoint. So at any point you can undo the prompts to get back to how the video looked before that specific prompt. And the next thing I'm going to tell it is based on the content of the video. Add a CTA screen at the end. Let's see how it does. Okay, that one took 51 seconds. Let's go to the very last scene of our video. Okay, just added a simple blank scene that says thanks for watching. With a placeholder, it's a start. And then I'm going to tell it. Add soft music throughout the video. Let's see how it does with sound. The music should be barely audible and complement the topic. Okay, it says soft. Barely audible music has been added took 14 seconds. Let's show the timeline. I'm not seeing a music layer. Let's go to the beginning. I don't think it actually did it. I think AI is lying to us. There is no music that's been added. Okay, let's see how it does at adjusting its own work. So I'm going to tell it. Go through all the text layers and change the font to Montserrat. So currently if I click on one of these layers, it's enter. Let's see how it does with adjusting what it's already added. Okay. It took 25 seconds and I can already see it worked. Check that out on this scene that I'm on those both text layers have been changed to Montserrat. If I go to a different one. Montserrat. Montserrat. Montserrat. Very cool. Now I'm going to tell it. Add punch ins to make the video more interesting. Okay. It tells me it's been added throughout the video. Let's see. Okay, here's one. So it goes from that view to zoomed in a little bit. Not drastically, but it did it. It jumped it to 115%. So pretty cool that it was able to do that. And there's a bit of a punch out. So right now I'm a bit zoomed in. And then in the next scene I'm zoomed out. In total, you can see that the recorder's been going for 24 minutes. I only briefly paused it when I stepped out to grab a snack. But the actual processing time of all of my prompts and me talking through it all is about 24 minutes and 30 seconds. And then what I would do now is just play through the whole video, put my human eyes on it, make sure that everything looks okay, clean it up. There's going to be a bit of timing, cleanup involved and just polishing some of the jump cuts, but pretty awesome. Overall, video editing is no longer a tedious task. It's now something you can do with plain English instructions. If you're a 40 plus year old beginner looking to create your own YouTube channel, check out the Descript Mastery School.
