Searchable transcript of How to Make a Micro Drama with AI (Full Guide) — Isa does AI (10:07). Search for a phrase, then click its timestamp to jump straight to that moment in the video.
Captions sourced from the original video on YouTube, published by Isa does AI. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.
00:00 If you've ever seen these micro-drama videos and wanted to make your own, now you can do that by just using AI. And it's not just one short video, it's a full 90-second episode with a cast, [music] a story, and a twist ending. In this video, I'll walk you through the entire [music] process from creating the characters to making a full drama episode by just using one tool.
00:18 And the best part is that the only thing that I'll need is just a few prompts. For all of this, I'm using OpenArt Director, and every prompt I'll use will be in a file in the video's description, so you can copy them and follow along step-by-step. Normally, every video is only a few seconds, and you put them all together after. But the director makes minutes of finished video in one generation with the characters, the voices, and the music already in it.
00:44 So this one's actually worth learning properly. When you open director from the menu on the left, you land in a chat box. The first thing I'll do is set it to guide me, so it checks in with me at every step instead of making the whole video in one shot. Then, I'll hit the plus to attach my character's face reference sheet and type in a couple of lines about what I want because keeping it that way is what gets it to suggest angles I wouldn't have thought of on my own.
01:09 That face sheet is the important part because it locks my look from the very first message and keeps me the same person in every shot later. This honestly feels more like vibe directing than prompting. It sets up the project, locks my face to my reference sheet, then gives me three different story openings to pick from. And all of them make sense, but each one goes into detail like describing the whole emotional baggage, which already makes it sound serious.
01:35 So from one message, I've already got three real story directions to choose from. But those are still just story ideas. The one I pick is where it actually starts turning into a real episode. So I'll go with the less obvious one here. I'll just tell it to run with number one, the overlooked heiress, where everyone thinks I'm just staff. [music] And the one thing I've learned from this is that you never treat that first pass as the finished thing.
01:58 You pick the direction you like, let it build from there, and react as it goes. That's the whole vibe directing thing again. I'm just steering, and it's doing most of the work. So, it [music] takes that one line and builds out the actual cast, and it's smart about it because every character is there to set up that one twist in the end. [music] I'm the assistant nobody looks at twice, and the twist is that I'm secretly the new majority owner of the company.
02:21 Then, [music] it gives me a villain, Julian, the board member who humiliates me in front of the rest of the guests [music] because a twist like this only works if there's one person whose reaction you get to watch. [music] There's the host up at the podium who makes the announcement, plus a few background guests so the gala actually feels [music] full.
02:38 And then, it lays out the whole 90 seconds across five quick steps. I'm walking through the gala carrying a coffee tray. Julian snaps his [music] fingers and treats me like staff. That's when the host taps the mic to introduce the company's new majority owner and points right at me. It even breaks down the reveal, my walk to the stage, and the final line that ends the episode.
02:57 And to get to this exact point with any other setup, you'd already be using at least two separate tools. You'd have one of them write the whole story out. Then, [music] you take that and break it into separate scenes by hand, turning it into shot-by-shot prompts so the image and video generators can actually use it. But here, it's doing both in the same [music] chat, the story and the structure.
03:17 So, there's nothing to copy out or reformat between tabs. [music] So now, I've got a full cast and a finished 90-second story to use. But all of this is still just names and lines on a screen. And now, I want to see if it can actually turn these people into something real. So, before it shows me a single face, it stops and asks how I want the whole thing built.
03:35 It gives me three quick calls to make, the aspect ratio, the resolution, and the length. Aspect ratio is the easy one, so I'll set it to 9 by 16 since this is all for vertical feeds. Resolution is the the that actually costs you Because 480p runs about 60 credits a second, 720p is around 110, and 1080p is about 230. So, it's the most expensive one. But, for a real live-action look, it's the best one to go with.
04:04 And the length I'll just set to around 90 seconds. So, I'll tell it the cast is perfect. Keep Julian exactly as he is, and let it move on. Then, it builds out the whole cast and the location, and I'll look at them one by one. So, first off, I go with my character. And on this close-up photo on the right, I feel like I look slightly different than usual.
04:22 But, it's nothing serious, so I'll go with it. Then, Julian, the board member who humiliates me, so he's the face you're watching when it all turns around. And he looks really, realistic, especially with the wrinkles around his eyes, so I definitely like that one. The host there at the podium, the one who makes the announcement. And again, it looks like a real portrait photo of her, so nobody would ever realize she isn't real.
04:43 And the last one is the gala itself, the hall dressed for the reveal with a clear stage and a podium. The whole place looks really realistic and official. [music] So, I'll definitely use it for the video later on. And what makes all of this work is that my face is built straight off that reference sheet, so it's locked from the start. And now I've got the whole cast and the gala built with my face locked to the one I started with.
05:07 But still, I'll need to make sure whether all of this holds up the second it starts moving. So, I'll give it the go ahead, and now I actually find out if it can pull the whole thing off. I'll just tell the model everything looks good, and that single confirmation kicks off the next big render. >> Over there, and keep it moving. >> Of course. >> The help stays at the back.
05:45 >> I was only passing through. >> Then pass through somewhere else. >> Ladies and gentlemen, thank you all for coming. It's my honor to introduce the company's new majority owner. >> Wait. The assistant? That's not possible. Miss I I didn't realize >> Thank you, Julian. That'll be all. >> I wouldn't say it's perfect though. [music] Watching it back, a few moments clearly need another pass.
06:40 So now I've got three inconsistent videos inside an otherwise finished video, and normally that would mean redoing the whole thing. But all I'll have to do here is tell it which videos didn't look good and regenerate the second, fourth, and fifth videos. And this is the part I really like about the director. You can step back in and fix individual videos anytime, no matter what stage of the video you're at.
07:02 It redoes the three broken ones with the same cast, the same parts, and the look while leaving the first and the third videos completely untouched. Then it joins all three back into the cut right where the old ones were. So I fixed it without ever redoing the whole video. The three are back in place, but the way they cut into each other still isn't smooth.
07:21 And that's exactly what the edit is for. The editing page has three tabs: shot, scene, and timeline. But I don't actually need to make any changes here, so I can go straight to adding captions. What I'll do here is I'll go to the timeline tab. I'll open the caption section and hit auto captions. >> [music] >> So So lays captions across the entire video on its own without me typing or timing a single line.
07:44 For something going on vertical feeds, you really want captions since so many people watch with the sound [music] off. And normally picking shots and captioning would mean a whole separate editor, but here it's all on the same page as everything else. Then once it's all cut together and captioned, I'll hit export, choose 1080p, and see how the final cuts look.
08:08 >> Over there, and keep it moving. >> Of course. >> The help stays at the back. >> I was only passing through. >> Then pass through somewhere else. >> Ladies and gentlemen, thank you all for coming. It's my honor to introduce the company's new majority owner. >> Wait, the assistant? Her? She's the assistant? [music] Miss, I didn't realize >> Thank you, Julian.
09:12 That'll be all. >> And honestly, it's looking really good now. The cuts flow properly, the captions are in, and the whole episode feels finished. And that's a finished 90-second micro drama made without filming a single thing. But a few details still feel off, like the moment where I appear to step over something that is not there. And the conversation with Julian somehow carries across the entire room, even though neither of us are speaking that loudly.
09:41 But overall, I am really happy with the result. But the best part is the limited time deal. Upgrade to advanced or above before July 31st and unlock up to 40% [music] off some of the best AI image and video models for the rest of 2026. [music] So, sign up through my link to OpenArt and build your first episode around one character [music] and an emotional twist in the end. Thanks for watching and I'll see you in the next one.