Let's Create AI Art & Thumbnails (LIVE)

To live streaming. This isn't my first live stream. This is my second live stream, but this is my actual first live stream, actually using obs. So I am, like, a couple minutes late, but I actually thought I was live. I was talking to you guys for a couple minutes and then realized I wasn't actually live. I think I'm live now. So if I am live, make sure you let me know in the comments, because I'm not 100% sure I'm doing this right. I'm actually learning as I go. So hopefully. Hopefully I'm doing this right now. Okay, so I see some people saying, look at that. He's live. Awesome. Hi. Okay, so I must actually be live now. I actually did speak for, like, two minutes while not actually live.

So again, this is my first time ever streaming using obs. Not my first live stream, first livestream using obs. So this should be fun. It's. It could be a total disaster because I don't go live often. When I make videos on YouTube, I usually have the benefit of editing. So what you're gonna notice is that I do tend to ramble, and when you watch my YouTube videos, I do a lot of rambling, and then I get to go and edit out all the rambling and just get to the point. When I do live streams, there's probably gonna be a little bit more rambling. So hopefully that's all right with you guys. I am gonna be kind of doing an ask me anything throughout, so feel free to pop your questions in the chat as I go. And every once in a while, if you see me look over here, that's because my chat is streaming right here, and I'll be looking over there to see if you guys have any questions as I go. Cool. Let's see. We're live. You're live. What's up, everybody? Tom Blake, good to see you here. Joel Moldowno. Rusty James. Awesome. Cool. We're good to go. So my sort of agenda is gonna be sort of freeform. Just hang out, have some fun. I want to generate some images, live, show you some of my workflows for creating things like YouTube thumbnails. Play around with some of the various AI image generator tools, test some cool prompts that I've seen on Twitter, and then the whole time, we'll be answering questions. And feel free to ask me anything, anything about AI art, anything about YouTube, anything about AI in general. If I know the answer, I'll do my best to answer it.

Alrighty, what do we got over here?

Awesome. So here's what I'm gonna do. I'm gonna jump into walking through how I make YouTube thumbnails. I've made some videos in the past, but this workflow has changed a little bit over time. I've added things in, like Photoshop generative fill. I use that as part of the workflow. Now I use Leonardo more often than I used to, so I'm gonna walk through the process. I don't have a ton of ideas for thumbnails, so if you guys have ideas for thumbnails that you want to see me try to generate, I'll try to generate them live on stream. It's always fun to challenge myself. I tell people when I'm in person at conferences and things like that that I feel like I can generate almost any image you can imagine with AI now. So it'd be fun to kind of put that to the test and challenge myself. I'm sure some of you will manage to stump me, and there'll be some things that I can't actually manage to generate with AI and we'll see how it goes. So one of the. One of the videos that I'm working on right now is a video about swapping your face in videos, right? So I've talked about stuff where you can swap your face in images, right? I've made videos about how to use Dream Booth to train your image in your likeness into Dream Booth. I've made videos about Insight Face Swap for Mid Journey where you can swap your face directly in Mid Journey. So I want to learn how to do it where you swap your face in videos. We make a whole video. I could take a TikTok dancer and have that dancer dancing around on video, but it's my face on that TikTok dancer. So, you know, some of the kind of fun, ethical, deep, fake kind of stuff. I want to make some videos around that in the future. So I figure maybe I'll make some thumbnails on this video for that. That's sort of my first idea for a thumbnail to generate. But again, I want to see some of your ideas. I won't be looking at the comments the whole time, but I'm gonna glance over every once in a while and read and see what you guys got for me. So to start, if I was making a video about how to swap my face on another video, and I don't know exactly what I'm gonna do for a thumbnail, I usually come over to our good friend Chat GPT here and ask for concepts for a thumbnail so I'd probably do something like.

So first of all, before I show you this, I do have some additional context over here in my custom instructions on ChatGPT. So if I click on custom instructions, you can see I put some details in here. I'm based in San Diego, California. I'm a full time YouTuber who makes my income through adsense revenue and sponsorships. My YouTube channel is all about the latest news and advancements in tech with a heavy focus on artificial intelligence. I can talk about AI, videography, cinematography, photography, tech and storytelling for hours. I have a goal of passing a million subscribers on YouTube and being the go to authority on the latest AI tech tools for creators, entrepreneurs and for productivity. So I have that context here. And then how Would you like ChatGPT to respond? Skip the formalities about being an AI chat model. Address me as Matt would. A long explanation is required. Give me a long explanation. If a really short and concise answer will do, I'd prefer that. So I have this additional context inside of my custom instructions here. So when I ask it questions and you see it respond to me back saying hey Matt or sure Matt, here's the response. That's why it's not just being creepy. It doesn't, it doesn't know me that well. I actually gave it that information so I wanted to show you that before I actually get in there because that might be relevant to some of the responses that it gives. All right, so I have an idea for a video. It's all about how to swap my face with existing faces in a video. What are some.

Concepts for.

Thumbnails for this video? Give me at least 10. Whenever I say give me at least 10, it always gives me exactly 10. But I for some reason always still put give me at least 10.

So while I'm waiting it for it to generate that, let me see what you guys are saying over here.

And thanks so much. Nick Vandems do you know any auto GPTs that don't require an open AI key? I don't. I actually don't mess with auto GPTs too much. I found that most of the auto GPTs that I've played with, they're fairly ineffective. Still, they, they don't automatically always get me to the end result that I want. However, I'm sure there are some auto GPT type tools out there where you can use things like llama2 and, and things like that that won't require an OpenAI key. I just haven't played around with them yet. It's something I do plan on exploring in future videos, but I haven't done it yet. Let's see. ChatGPT for Enterprise. Did you see the news? Yep, I saw that. Seamless cat patterns. We can play with that.

How to turn yourself into a baby. All right, all right, we can play with that one.

All right, cool. So let's see some of these ideas that it gave us here. So split screen your face on one side and the swapped face on the other with a dividing line before and after. Show a side by side of the original video frame and the swapped face. Video frame face in progress. Show a half swapped face like your face is morphing into the other. That one's kind of cool. Zoom and swap a close up of a face in the video. Then another close up but with your face swapped in. These are all kind of the same concept. Timeline Represent the swapping process with a video timeline at the bottom. Marketing marking when the face swap occurs. Multi swap reaction. Tech overlay Swap button. Dramatic change. I think the face in progress one is kind of interesting here.

So let's go ahead and try that one. So show a half swapped face like your face is morphing into the other. So what I'll do is let's go ahead and jump into Mid Journey. I actually created a brand new little room just for this live stream. So it's starting fresh. And let's prompt an image of let's say.

A man doing a dance for a TikTok video. Let's see what we get. I'm going to make it aspect ratio 16,9 and let's see what it generates there while it's generating that. Let me take a peek over here. How can we upload Driver Sensei. Hey, thanks for the 5 bucks. How can we upload an image and get it to output something similar without much changes to it? Well, if you're using something like stable diffusion and automatic 1111, that's pretty eas. If you're using something like Mid Journey, that's a little bit more complicated. But they do have inpainting which makes that easier. So we'll get into some of that in this video as well.

Let's see.

Why don't you have to upload specific workflow stacks? Jib jab. Who is AI Larry? I'm not sure.

All right.

Face swapping for videos. Yep. There's also so face fusion for face swapping. There's also Roop, which is what I've been playing around with a little bit.

So many comments. All right.

So we've got. We've got a few TikTok dancers here. These are. These are interesting.