Ilan (00:00)
And I am too lazy to come up with the entire prompt on my own, so I'm just gonna tell it using the styling can you create a prompt for the scene? The bog the bob

David (00:10)
The bog darks at the mailman.

faceless explainers are a great way to engage your audience and help them actually learn something at the end of the day.

Ilan (00:32)
Yeah, that's right, David. They are really easy to create, and we've been making them for the last few weeks. Here's an example.

So let me walk you through how I got the first one built and how I made it a repeatable skill so I can build them over and over again.

Alright, so you have something that you want to explain to somebody. And you've recorded yourself explaining that.

So you want to have some engaging content.

AI is really great at creating images that can be stitched together and create a video.

I'll walk you through the steps.

First step, find similar videos that you like the style of. Here's what we did for that.

Alright, so this is a YouTube channel. You see that they have 172,000 subscribers and 29 videos. And this top video here has 7.8 million views.

The style is really simple stick figure images, maybe something that we'd be inspired by. So first thing we're gonna do is grab some screen captures. you can grab them from different moments in the video.

And I'd suggest that you grab like eight to ten of these.

David (01:46)
This is good. You are grabbing a variety of different scenes

Ilan (01:53)
For this entire

Explainer, I am going to be using the ChatGPT app. It used to be called Codex. And all you need is your basic monthly subscription to ChatGPT. And the first thing I'm going to do is I'm going to load in those screen captures that we took.

David (02:08)
Now I notice that you have it set to dangerously accept full access to everything. is that required for this?

Ilan (02:16)
You absolutely do not have to. I'm gonna keep it on for the purposes of this video so that I don't have to approve every five seconds while it's going through this flow.

Okay, so we uploaded our images into ChatGPT, and we're gonna give it a prompt telling it to analyze these images visual style for use in an AI image generator, and get a positive and negative prompt as well.

David (02:39)
And notice that you're using GPT5.6 Terra. Is that the right level for this?

Ilan (02:44)
you don't need to use GPT5.6 Terra, but we're gonna use multiple turns in the same conversation. So you want a model with decent context window.

Okay, so here just after a couple of seconds, we got our positive style prompt and a negative prompt. So the next thing we want to do is just test this out on a single image. Let's make sure that it works for an image generator.

Alright, so to test this out, we're gonna get it to create an image and we're gonna give it the style prompts that it gave us previously. Let's see what happens. The cool thing with using Chat GPT for this is it has a great image model, GPT image two.

already built in

David (03:22)
I'd imagine this will also work with Gemini then, given that they can also generate images.

Ilan (03:27)
Yeah, exactly. the one it doesn't work well with is Claude.

Alright, so here we got our image. It's got that simple hand-drawn style that we're really looking for. So now we are going to test this out on a short clip.

Alright. So now we're gonna go full Megilla. We have an existing transcript that comes out of our recording tool. This is for about a twenty-second segment. And we are telling it to generate images from this script.

It is going to pull the timestamps out of the script and create an image for each timestamp.

We've given it the styling prompts and

We've also told it to create a folder and save each image to that folder named appropriately.

Alright, here we are. We've gotten the model to create images for this 20 second audio file. And the next step is to get it to stitch it all together.

So that's what we'll tell it to do. And we're specifically telling it to use this FFmpeg Python library. it's a video editing.

Python library.

all right, so there we go. It has created our 20-second video.

this whole process took about 15 minutes between the image generation and stitching it together for the video. It doesn't take that much longer if you do a longer video. You know, if you do a one-minute video, it might be 20 minutes. So you're not looking at a linear scaling from here up. And

David (05:01)
And one

of the good things about the using the GPT image model is that it's pretty good with text.

Ilan (05:06)
That's right. And the last thing that you should do is to turn this into a skill so that it's something that's repeatable for you. So we're gonna turn this into a repeatable skill now. We've told it to have two entry points, either give it the text file or an audio file and have it figure it out.

And it's going to just run this itself. And because we have that skill, we're actually gonna share it with you. So we'll post the link below and you can click that and get the skill and you can create your own faceless videos.

David (05:35)
Awesome. This is this is great. thanks for walking us through this, Ilan and I hope to learn a lot more through faceless videos now.

Ilan (05:43)
Absolutely. All right. Thanks everyone. Subscribe, give us a like, helps us expand, and let us know what else you'd like us to explain.

David (05:52)
All right, we'll see you at the next one.