Gen AI University

← Video Library · Content & Video

Create Video Content 5x Faster with This Strategy

Published April 3, 2025 · 250 views on YouTube

Key takeaways

How can AI cut video production time for content creators?

Steve Kerry, a member of Gen AI University’s mastermind, combines two tools to compress his content pipeline: Descript for editing and adding transitions without traditional video-editing skills, and MindPal for turning a finished video’s transcript into a complete YouTube publishing package (titles, description, keywords, thumbnail captions, newsletter, and social posts). He walks through both tools live using a real client project, a 21-day sleep challenge that requires a video, a newsletter, and social posts every day.

How does Descript speed up editing without traditional video-editing skills?

Descript lets you edit video by editing the transcript text instead of a timeline, and Steve uses a few specific features to remove nearly all manual editing work. First, he runs Underlord to automatically detect and strip out filler words (“ums,” “uhs,” and similar) across the whole recording. Then he enables Studio Sound, which cleans up background noise and improves audio quality even when a guest is recording on earbuds rather than a dedicated microphone.

For visual polish, he applies Descript’s built-in layout packs (customized with his own fonts and branding) to add lower-third name/title cards and switch between a full-camera view and a screen-recording-with-speaker-box view. The key feature he highlights is smart transitions: Descript automatically handles the transition when you switch from one layout to another, so no keyframing or manual transition work is required. He estimates the roughly 23-minute video shown in the demo doesn’t take long to edit once you’re familiar with the workflow, and that the process can be documented into a standard operating procedure for a team.

What is MindPal and how does it turn a transcript into a full YouTube publishing package?

MindPal is an AI agent-based workflow tool. Steve describes a “workflow” in MindPal as a series of independent AI agents, each handling one discrete task, chained together so the output of one step feeds the next. He builds a new workflow by writing a plain-text description of the steps he wants and having MindPal generate the agent chain from that description automatically.

Once a video is finished in Descript, he exports the transcript, pastes it into his MindPal workflow, and the agents run in sequence: generate YouTube titles, pick a winning title, write the YouTube description, generate SEO keywords, generate thumbnail captions, generate an email newsletter, and generate two styles of social post. Because the export runs against the transcript automatically, most of the manual work left is copying, pasting, and light editing.

What are the steps in the MindPal content workflow?

  1. Export the finished video’s transcript from Descript and paste it into MindPal.
  2. Generate YouTube titles (Steve trains the agent to produce five titles, review them, generate five more, and recommend one) and select a title.
  3. Generate the YouTube description, which pulls in the winning title, call-to-action links, and matches example descriptions Steve provided as a knowledge source.
  4. Generate YouTube keywords, formatted as a single comma-separated line for easy pasting.
  5. Generate thumbnail captions to choose from or mix and match.
  6. Generate an email newsletter for subscribers.
  7. Generate two social post variations: one natural-sounding, one caption-style.
  8. Copy the finished title, description, and keywords back into Descript’s YouTube export screen, add a thumbnail, and schedule the video.

How does fine-tuning make the AI output sound less generic?

A freshly generated MindPal workflow is generic by default, it has no information about brand voice or desired output format. Steve fine-tunes each agent individually by attaching a knowledge source (an example markdown file showing what a good description or newsletter should look like), selecting a saved brand voice, and choosing which language model powers that specific agent. In the video he uses Gemini 2.0 for the description agent and GPT-4o for the newsletter agent, noting he often prefers Anthropic’s models for newsletter writing specifically because of how they write. He flags social post copy as the hardest output to get right, saying it tends to still “sound like AI” and requires the most additional training.

How long does this workflow actually take in practice?

Steve estimates that once you’re in a groove, running the MindPal AI workflow itself takes about five minutes, plus a few more minutes for copying and pasting content into place. Video editing in Descript still takes some time on top of that, but he says the output from the AI workflow is typically 90-95% ready to publish with only minor edits, and that batching multiple videos through the same process lets you produce a lot of content quickly.

Full video transcript

How’s it going everyone? Darby here with GenAIUniversity.com, and in today’s video I’m with Steve Kerry, a member of our mastermind, who’s going to be breaking down a super simple, highly effective content production workflow to help you create better content faster and get it actually published. Let’s rock and roll.

Hello everybody. My name is Steve Kerry, and I am so pleased to be part of the Scale with AI Summit for 2025. I wanted to share with you why I am so excited about the promise of what AI is bringing to us. My experience throughout my career, I’ve always been involved in some type of high-tech work, starting with help desk work at the dawn of Windows 95, replacing mainframes, working in ISPs, working in robotics, working in Internet of Things. One thing that I know to be true: technology can be intimidating for so many people, and I’ve always recognized that language around technology and how it gets used is a blocker for so many people to be creative.

The big reason I am so excited about AI is that I think this is one of the technologies that allows all of us to access the creative part of us inside, and we’re able to use the tools in ways that are so much more intuitive, easy, and fast. And I’ve seen time and time again that so much of the tools out there are difficult to use and usually require somebody with tech knowledge or some type of experience that is difficult to learn. If you just have a job and you want to make a video, it is so complicated to do video editing. And that’s what I’m going to share with you today. I’m going to share my entire process of how I use AI to accelerate my content production for my YouTube videos, for me and for my clients. The result is it saves me a lot of time for my work, and it saves my clients a ton of time for the work that they do, and they can now focus on their message and the audience that needs to hear it. And I think that’s a wonderful gift.

So I hope you enjoyed this video, and I hope you’re inspired by what you could potentially do by using tools just like what I’m about to show you. And just know that AI is there to help you get your voice out to the people that need to hear it. All right, let’s get started. Today I want to show you three things, and this has everything to do with creating content quickly and getting it published. First I want to show you Descript and how I use Descript in my content workflow for both myself and for my clients, and how this not only saves time but allows you to be more creative and not worry about the tech, the mechanics, and all the details that make being creative so difficult. The other tool I’m going to show you is a tool called MindPal. It’s an AI agent-based software, and the one thing I’ll say about it is it makes it really easy to create automated workflows that combine multiple steps in a way that is really easy to use and produces really good output. And then finally, I’m going to put this all together and show you how I’m taking a current video I’m working on, running it through the examples I’m showing you in this script, taking that transcript from that video, putting it into my AI workflow in MindPal, and then showing you how I generate the necessary details to get it published on YouTube really quickly. So there’s a lot here. I’m trying to get it down into about 20 minutes, and it actually doesn’t take that long, but I’m trying to show you a lot of things all at once. So here we go.

Right now on my screen, you can see I’ve got Descript pulled up, and just the other day I had recorded a video showing the work process I’m using in MindPal, a very similar video to what I’m doing today, but showing a process I’m using with one of my clients, and she’s given me full permission to talk about the project. So I’m showing in this video on the screen how I use MindPal agents for this 21-day sleep challenge. And the reason this is significant is because each day of this challenge we’re producing a video, we’re producing a newsletter, and we’re producing social posts. So there are a lot of things that go into the mix, this client work, and I showcase that all in this video I’m editing right now.

What I want to show you is a few things in Descript that make it easy to create a high-quality video that’s interesting to watch, sounds good, looks good, and has all of those cool transitions that you see in some of the higher-end videos. Descript has figured out how to make those transitions really easy to use for people like me and you who don’t want to spend all of their time doing video editing. And that’s what I love about this. So let’s get into it. What you see right now is my initial video, this is how I start out, and my workflow is this: I always make a working copy of my initial video so I can always go back to the first version. I’ve made a working copy of the first version, ready to go with the Scale with AI Summit we’re doing right now.

If you haven’t used Descript before, the one thing that made it popular is that you can edit videos in the same way you would edit a Word document. You see the script on the left, the video on the right, and down here is the timeline, which is usually what you’d see in most video editors. The secret here is I rarely use that timeline. It makes it really easy to do editing just by editing the words. So the first thing I do is use an AI tool called Underlord. I won’t spend too much time on this, but I just want to show you that you can go in here and identify and remove all of the filler words. You can see when I made this video earlier there are a lot of ums and o’s and things like that, so I get rid of them, and there are 23 of them. You can see them all right there if you wanted to listen to each one, but I’m just going to remove all of them, and just like that they are gone.

Now the other thing I always do at the very beginning, before I start doing a lot of editing, is enable a feature called Studio Sound. This makes the sound so much better, especially if you have a decent microphone, but even if you don’t, I work with clients who just use earbuds, and when you enable Studio Sound it gets rid of all the background noise and just sounds better. So I do that right away.

Now we’re at the starting point, and what I want to show you next are a few features that make it really easy to add neat transitions. You’ll see I’ve got a screen recording, and there’s me in a box in the lower left-hand corner. What I want to do is introduce myself, so I highlight this bit of text, go up to Scene, and up here they have a feature called layout packs. Descript has several built-in layout packs, and I took one of them and did some edits to change the fonts and some of the branding so it’s more in line with me. What I want to show you here is that you can choose a feature called Speaker, and it applies this template just for the part you highlighted. When I go play it, you’ll see I’m already saving hours of time. It brings up a speaker name, and I put in my name, and here is where I’d put in the title of the video I’m doing. That’s a lower third that just pops up, and you saw how quickly that happened. Key here is they’ve built in the transition.

In this next section I add another layout, Camera, so I’ll have the lower third and this beginning part, and then it goes to full camera. This video is part two of building an AI-based workflow in the tool called MindPal, so I’m not ready to go down into the screen recording yet. I want to put some focus on the MindPal part, so there’s another layout called Media Layout, where it shoves the speaker camera view off to the side and gives an option to put a screen overlay on the left side. I go up to Project and drop in the MindPal logo I’ve already got. So now it reads “workflow in the tool called MindPal, the first video,” and then I return it right back to Camera. Now we’ve got this really quick sequence where it’s me in the tool, and you’ll see on the left side we’ve got four scenes.

Now, as you go through the whole video there are a lot of those, but the part that makes this so compelling for people like me, and hopefully you, is that I spent zero time working on transitions. I didn’t make any keyframes or any technical things like that. The feature Descript put in is called smart transitions, it just knows that when you want to go from one version to another, it does the transition. Just for fun I’ll put another scene break to show the full screen of this video. It would take too much time to go through the entire editing process, and you’d be bored out of your mind. What I can attest to is that this video, in its current form, is roughly 23 minutes, and it doesn’t take that long to go through all of this. You get really quick at doing it, and once you’ve done this a few times you can start training other people to help, and put it into a standard operating procedure.

Now, just like in a cooking demonstration, I have a version I already did, and I’ll show you a couple of things I did at the beginning of the workflow. This video is part two of building an AI-based workflow in the tool called MindPal. The first video showed how easy it is to create the workflow, but we didn’t do any fine-tuning of the agents or the content. This gives you a complete example of how in-depth the entire video is, and when I’m done I have a final version ready to go.

Now I want to show you how I take my ready-to-go version and build out all of the content that goes with it, so I can publish this to YouTube really quickly. The first thing I want to show you is MindPal. I’m going to demonstrate how easy it is to create a new workflow. A workflow is just a collection of independent agents working in a series of steps, each executing a discrete task, and MindPal makes it easy to build one. My workflow is this: it starts with using the transcript, I create YouTube titles, I pick the YouTube title I like most, then it creates a YouTube description, YouTube keywords for SEO, a caption that goes into my thumbnail, an email newsletter that goes out to my subscribers, a natural-sounding social post, and another social post that’s more of a caption style.

How do you get here? This is where the magic of MindPal works. Let’s create a new workflow really quickly. The beauty of creating workflows in MindPal is you can use AI to generate them. I’ve already got a description of the workflow I want, since it’s easy to do, I just copy and paste that series of steps, which I put into a text file, and paste it in, then say generate with AI. You can see it’s doing what I want: manual step, generate video titles, then select a winning title, another manual step, then it writes the YouTube description, generates keywords, generates the captions for the YouTube thumbnail, email newsletter, social post one, social post two. It did it exactly the way I described. I’m going to call this “demo” so I can keep track of it later.

The beauty of how this works is that this is a starter workflow, MindPal goes through and creates each agent for each step. This one has an agent called Video Title Generator, and behind it MindPal created the prompt in the background for this AI agent, then created the output format, all based off the few text keywords I put in. There are additional instructions here because MindPal is telling it to use the input from the previous step, transcription. The way you do that is by using the “at” symbol to get the option for the transcription input, and as you go down to later steps I’ve got other input from previous steps built in. In the task section you can give it additional information to get very specific.

The main goal of showing you this is I wanted to show how quick and easy it was to create a workflow. Now, this workflow is generic, it’s not fine-tuned, it doesn’t have information about what the output should look like or my brand voice. But in this finished one I’ll show you one step to illustrate what I mean. I’ll use the Create YouTube Description step to showcase this. Here’s the task information: I’m telling it to use the winning title, the transcript, the user input from context, and call-to-action links. Up here in the agent there’s background information, and additional information about what should go in a description. Behind all of that there’s a knowledge source, an example of what the YouTube video description should look like, in a markdown file. Then I can choose output settings and select a brand voice, I have a brand voice set for my YouTube channel. The most powerful part of all of this is you can choose which model you want to use. In this case I’ve got Gemini 2.0. For some other agents, like generating an email newsletter, this agent uses GPT-4o, and I’ll switch these back and forth. Often the newsletter one will be Anthropic because I like the way it writes email newsletters better.

Now you get an idea of how easy it is to create a workflow, and all you need to do is go into each agent and start fine-tuning it by adding knowledge sources, choosing which language model, and all the things you want to do to fine-tune it. Once you’ve got it done, all you have to do is click New Run, and it’s ready to go.

Let’s go back to Descript, and I’ll go to export, choose transcript, and copy to clipboard, you don’t even need to save it to a text file. I’ll enter the transcript. For this version I’m not going to put any call-to-action links, but if you have a group or an affiliate code or anything like that, you could put that in, it’s an optional field. Now I just hit continue, and the first thing it’s going to do is generate a bunch of titles for my YouTube video based on the transcript. I like to have a number of different titles, I’ve trained it to generate five titles, review those, and generate another five, and it makes a recommendation. It’s got “one hour to minutes AI content workflow revealed.” For the sake of time I’m going to use this title, it may not be the best one, but I’ll use it today to go fast.

I copy that title and paste it into the next step, which is user input, and everything from here on out is automated. It goes and creates the YouTube description, incorporating my title, and it’s got everything I want to see, matching the examples I put into my knowledge files, with an actual call-to-action link for MindPal, which I’d trade out for my affiliate link, and it’s got notes about being a hypothetical link. Then it has timestamps. This probably requires a little editing, so let me show you how quick and easy it is to edit this, I just hit edit, swap out the placeholder link for my affiliate link, and the timestamps look good.

I’m ready to go to the next step, YouTube keywords. The other thing I trained this agent to do was output this all in a single line, comma separated, so I can copy and paste when I need to use it, and that looks good. Now I’ve got thumbnail captions for my thumbnail image, these are all just examples, and it saves me a ton of time because this is one of the jobs I dislike. I usually mix and match from different sets and make a version I like best, and there’s a recommended version. I’d like it to emphasize MindPal, because when I put this up on YouTube I want people searching for AI agents and MindPal to find it and see how to make an AI workflow really fast.

Now on to the newsletter, I’ve got this dialed in, and it matches the examples I’ve used before. Same thing here, if I need to edit it I go to the edit button, click save when done. Then I can copy that and go right into my email sending program and drop it in. Now I’ve got a social post, and I think I must have messed up on this agent, it looks like it’s missing the input, so that’s a part I’d need to go back and fix. Then here’s a social post in a caption style, and it did an okay one. I think out of all of this, writing good social posts is probably the most difficult and most challenging part of getting AI agents to do what you want, because they always sound like AI, so I find I have to go back and train these even further.

Let’s back up a minute and look at what we have. We have all the content I need in order to publish this video. The way I do this is go back into Descript and start loading it in. I grab my title and change the title here to that one, I grab my description and copy the whole thing, then I hit export and choose YouTube, clear the old description, copy in the new description, and it holds without breaking. I’ve got my keywords, and I drop them in. Now I’m ready to go, I have everything I need to publish this into my YouTube channel. Once that’s done I can go back and add in the thumbnail and schedule a time for it to be released.

Why this works so well is you can see I’ve got all the parts of this job that take time, the copying and pasting, moving between screens, minimized, and I get really good content that’s 90-95% good enough to go with minimal editing. Then I can go to the next step. This is easy to delegate because there aren’t too many decisions to make outside of choosing a title, and when you have a team of people who understand your brand voice and how you want things to look and sound, it’s really easy to get this level of editing done. This saves so much time, and when you combine this with the editing features of Descript, the ability to make transitions and all of those features you saw, you don’t need to be a technical expert to create good content, to communicate what is in your mind, and the thoughts, emotions, and feelings you’re trying to convey to your audience. That, to me, is what makes AI so powerful, because it opens up our ability to be creative in ways we couldn’t before, because we didn’t have the time and we didn’t have the technical capability. AI unlocks all of that for us and opens up a whole new world to do more, to create more, and to share more with people who need to hear your message.

I really hope everything you saw here today was useful to you. The last step I need to click is the export button, because that sends this video up to YouTube, and I’ll drop in a thumbnail based on the captions I showed, and then I’ll be done. This video, when finished, is going to be right around 20 minutes. When this is all said and done, this whole process, when you get in a groove, is about five minutes to do the AI workflow, plus another few minutes for copying and pasting. It does take a bit of time to do the video editing, I don’t know how you get out of that, but it’s really quick once you start batching these up, you can get a lot of videos done very, very quickly.

I look forward to your feedback, let me know any questions you have in the comments, and I hope this inspires you to take the next step in your journey to create more video content using AI agents to really hone in and dial in your message and get more done. All right, that’s it for now, so happy to be here, and have a wonderful day.

Watch on YouTube

Related tutorials