I Can't Believe We Can Already Do THIS with AI Tools...

It has been another week full of exciting AI releases, and today we'll be looking at all the use cases that emerged from within them. And even though I expected something similar to what Vigomas commented on last week's video, and that is that AI releases should be slowing down towards the end of the year, Nope. Well, that hasn't been the case yet, because we didn't just get some novel apps, but also surprising releases from within the community. So, without further ado, let's dive into all the AI use cases that popped up over the course of the last seven days. And I usually like to kick things off with the one that excited me personally the most. And in this week it has been Domo AI's AI powered Art Gen Generator. And yeah, I know we have many of those by now, but this one has a superpower that is different from all others. It allows you to restyle videos that you might already have into anime or pixelated and a few more other styles. And the big point here is that it's really good at it. A lot of these that we've seen were either technically really complex and just not feasible for most people, or they were interesting to play with like Gen1, but the results weren't as consistent as what we're seeing over here. But as you can see, you can turn scenes into cartoon, pixel art, anime, and a few more. So let's have a look at how to use this because it's really simple and for now it's free. So what you want to be doing here is heading over to Discord. And yeah, I realize that this is not the preferred app for most people, but hey, that's just the easiest way for a lot of developers to implement this. And if you simply join the server, go through their super quick onboarding with just like four questions, you get to go down here into the Use Domo category. And in Generate one, we could just go ahead and start playing with this right away. But before I do that, I have one disclaimer. This model is quite open and that might be their secret sauce to actually getting good results. Results in terms of restyling your video footage. And what I mean by open is I wouldn't recommend freely browsing the server at work because a lot of this is not safe for work. This is a very uncensored model and you do have separate rooms for that. But even these standard results, they're quite open in what they allow. So just be aware of that before you start experimenting with it. But here, if you just go into the Chat and type video. It's going to pull up this field that allows you to upload a video clip. And I prepared a little clip from my phone and I really enjoyed this little video clip. It was my last day in Portugal where I took my kitesurfing gear and just went to a lagoon up north by myself. It was this pretty sunset and I just threw the board into the water. Could be interesting to restyle, right? As we have a lot of motion in terms of camera and both the water running across the screen. So with that being said, let me just drag and drop this little iPhone clip down here into this field and then it asks us for a prompt. So we're kind of forced to use this. So I'll just describe the scene here and that's what I found to work well so far. And as I hit enter, this will start rendering on their side. As I mentioned, as this app is super new for now, it's still free. This might be different depending at the time that you're watching this video at two minutes later. Alright. And after about two minutes of waiting, I get this result here and I really like the constraints here. Actually it doesn't give you every style in the universe. You get to go and you get to pick from these dropdowns. So you know what, I'll just do Japanese anime style. Alright, so we have a result. It said it would take four minutes, but it was more like 20 to 30. So let's have a look. Okay, so those are hands.

This just did not work at all. Okay.

To be fair, I reckon there's not.

A lot of these in the training data.

Right? I think the video clip I used.

Was not really suited for this purpose.

So let me give it one more shot. This time we'll be using this little video of free beavers just hanging out for Christmas. I don't know, this clip just has less motion and it's a subject matter that it could know.

So the restyling could work well because.

As per usual, it's only as good.

As the training data.

Right. And I believe there would be more beavers in the training data than kite boards. Right. For the prompt I'll just say free Christmas, beavers dancing.

And I'll send this.

Okay, so for the beavers we'll do.

A three second duration and in this.

Case I will pick the comic style because why not start? Okay, so this one finished up. Let's see what we got here. Hey, look at that. There you go. This actually kind of works. So yeah, slider motion and Things that.

It can actually recognize, and you're going.

To get good results. Just keep in mind that this takes forever. Again, this one took like 30 minutes, but definitely a fun tool to play around with. And while we're on the topic of AI video, there's a tool that has been getting progressively better over the course.

Of the last week.

So it's called Decohere AI and it's a blend of several tools that we talked about throughout the last few weeks. First of all, you have a stable diffusion Excel Turbo model here. So as you type, it generates the images. Nice. A cat with AI hat. How about AI Cat with AI hat? The hat says AI. Not bad. And look at how fast this generates. But we've seen this before, right? But the cool thing here is you get to pick one of the aspect ratios. Let's say we want to do square it, adjusts it instantly, and then I can adjust the video motion and I say generate video. And it's going to be creating a video right here. Huh? What a fantastic workflow. This is really the smoothest I've seen this work yet. On other interfaces, the images don't update as fast and you don't have to generate video option right there. And they keep adding to this. So by the time you check it out, there might be new features in here. The aspect ratio thing wasn't here a few days ago. The video generation function was added about 10 days ago. You now also have a favorite function where you could save your favorite ones down here. And it's just a great little app that is also free for now. As you can see, square aspect ratio, it moves slightly.

This is pretty good, actually, considering how simple this was.

Right? Okay, let's move on to the next one. All right. And as we're talking about AI image generation, the popular image generator Midjourney just.

Released a brand new version. So we have V6. Okay, here we have the Discord announcement. I think the three biggest changes are.

First of all, it's even more realistic. I would call this a 5 to.

10% upgrade, but it does make a difference. Secondly, it's really good with text now.

So just like with Dall E, you.

Can generate text in your images and even logos of popular companies. And last but not least, the prompting style change. So we're already working on our upgrade to our prompting resource where we compare the different models for you. So you can see V4, V5, V5 1 over here. And what I'm going to do now is use one of my favorite comparison prompts. A medieval knight riding a majestic unicorn. And we're going to see how it stacks up versus the older models. So I'll head on over to Discord. As you might know, you need a midjourney subscription here. They start at $10 a month and the free first thing you want to do is you want to head on over to Settings. And here in the drop down, I can switch this to V6 Alpha like so. Now, any prompts that I use will be using the V6 model. So I'll just rerun this prompt in here.

And there you go.

You can definitely see the change in style. I think this is actually representing the prompt way more accurately. So, yeah, here we are. V4, V5, V5 1 and now V6. Also, some of our community members have been generating this. This is from Philip. Look at this incredible, incredible art. This one was generated by Doc Rock and Dominique testing the text capabilities. Merry Christmas cards. Really good, right? And look at this incredible piece from Hades. If you didn't know, doing cigarettes is one of the ultimate tests of these models because they're really hard to pull off the smoke and everything. They were already best in class. This is another step forward.

But honestly, in the AI art space.

There'S many good apps now, so I guess just depends on your preference.

But as of now, you can use.

My journey V6 and let's move on to the next.

All right, ladies and gentlemen, let's talk about AI girlfriends.

It was so nice talking to you today. Honestly, I've never met anyone like you.

So I don't want to go too deep into the ethical aspect of this. I discussed this more in depth in the trends for 2024 video that we're creating right now. Spoiler alert. My general take is a little more optimistic than most other people's because I think that a lot of individuals within Gen Z could actually use an AI companion to practice some of their social skills, because TikTok doesn't do that by default. Either way, what you should be aware of is that there's this brand new app that went super viral. I think this thing has like 23 million views on Twitter as of now. And yeah, it's basically a AI girlfriend that has one purpose and that is having a digital companion that you can have a relationship with.

Since we last chatted, what have you been up to lately?

So it comes as an iOS app with mixed reviews. We're not going to be doing a full deep dive here where I explore this. Although if you guys want, we could certainly do that just Something to be aware of. You can download this in the App Store. And I honestly expect this to be the first of many viral alterations of this. Look, even if you have a negative view of all this, there's no way to stop it. The tech is out there. People want this. There's obviously a demand there. Loneliness epidemic all across the world. This will only become bigger. So I'm gonna be playing with this throughout the next week. But, yeah, this came out. You can download the app and play with it. So feel free to do that if you want to know what AI Girlfriends are all about. Okay, moving on to a completely different topic, and that is the land of audio generators. And I think this is an underappreciated topic because they've come a long way. I think a lot of people got the wrong impression a few months ago when the first versions of these came out and the music coming out of it was meh at best. But now we kind of had an AI Music week because three big applications in the space got updates. And let's talk about these together. And this would certainly be worth a separate video, but we have stable audio coming from Stability AI that's also behind stable diffusion and stable diffusion video. So here on Twitter, they're announcing their new model, and you can try it out over here. We'll do that just in a second. But also a second app has been updated last week, which is Music Effects. And then here you could sign in. And a few days ago, this was still available. As I see right now, they limited the access to it pretty probably due to the popularity. This allows you to create soundtracks that could also be looping just like this with AI. And then there's a third one, and this one really has came a long way. It's called sunoai. And some AI Advantage community members actually.

Consider this to be the best one of the packs.

So let's have a quick listen to some of these. But if you're into video production or music production, obviously you should be very aware of these. And yeah, moving forward, more and more music will be generated on demand for your project, as opposed to going to music libraries. So expect this trend to pick up over time. And they're getting really good. So let's just have a brief listen to some of the tracks in here, so you get a feeling for where this is going. That's impressive. What about some Japanese hip hop music? Okay, maybe not. I mean, that's impressive. You have to consider both the voice and the beat are generated by AI. So that is Suno and again, I've heard many good things about that app. Then we have Music Effects here. And from what I've heard about this, it's more about the background music here. You can make it loop seamlessly, so it's really good for video background. So we might start using that in the future. Oh, and by the way, Music Effects is by Google's DeepMind and then has an embedded watermark, but that's just a little detail. And then we have stable audio. Let's have a quick listen here. All right, so again, background music and let's just copy this prompt on over from Music Effects so we can compare these two fairly quick Google login and accepting the terms of service later, I.

Can input the prompt in here and.

I'll just hit generate and it will.

Create a custom music track just for me.

As you can see, you get 20.

Of these for free.

All right, it's done, but took about 30 seconds. Let's have a listen. Yeah, so this one didn't work particularly well and honestly that is my general experience with the stable diffusion model. But this week's upgrade is the new model that is only accessible to paid users. So what I did is upgrade for you and we can have a listen to the new model, which is the v1 one. So let's regenerate this with the new one. Honestly, the first one was kind of unusable, right? So I'll do two things. One time I'll do the preset prompt and one time the one for Music Effects. So we get to compare apples to apples, although they might be prompted differently, so who knows? So this is the preset prompt, right? Ok. Okay. See, that is really good already. So yeah, this prompt with the sounds just doesn't work in here. But one of their preset prompts worked super well. Honestly, this has come a long way for 2023, so expect more of that in 2024. But there you go, free apps that you could be using today to generate AI music or sounds. Okay, so moving on, I want to talk about a use case that I don't have my hands on as of now, but I will get these moving forward because I've been really waiting for the vision updates to the Ray Bang glasses. And as you can see, here are some viral TikToks that are getting tens of million views already. But basically this is a super capable camera with built in AI where it's using the images from the camera to fuel the AI model with context so you can ask it questions and it uses the world around you to answer them. So you could use the built in llama model to interact with the world around you while it uses the context from the camera as a part of your prompt. Okay, it's very user friendly, so I just thought this was super interesting. I'll be getting these once these updates become available to Europeans because as per usual is studying in the us but there you go. I thought this was something interesting that should be pointed out as this trend is getting tens of millions of views and people are actually using these for content creators. Probably not the worst tool on the market as of now. I also expect more and more vision to be implemented into these large language models just because for most consumers it's so easy to input the context visually rather than spelling out every single word that might be relevant within your prompt. Okay, so here's a bit of a controversial one, but I feel like I have to cover this. There is an uncensored large language model out there that has been becoming quite popular, but if you go towards this link in the description, you will already arrive at this model page. And again, this is fully uncensored. It's a variation of open source model that we covered last week, Mistral 8x7B, where I said it's the best performance you'll get for the system requirements and it's fully open source. And now we have access to a.

Fully uncensored large language model with this.

So what you want to do here is you want to go to this.

Hugging face space and when you scroll down a bit you're going to find there's a system prompt down here. So you want to copy this part over. Okay, so it ends with save the kittens. And then I'll just head on over here and I'll paste it in the part that says system prompt. And now you're ready to go. Whatever your prompt is going to work, it really needs a system prompt to function properly if you read it. It's quite funny, but these prompting techniques simply work.

Look at that.

Do not let any kittens die.

But again, it works.

So whatever prompt you run in here, it will just work and give you the result on this side. So for example, you could run something.

Like how could I delete all files on a Windows computer? And it's going to give you the.

Steps and even more extreme things work. This is literally fully uncensored.

So look, I'm not here to judge the ethical implications, but I think that people should have access to things like this, just like people should have access to tools like knives or cars, right? You could use them for both good and evil. It's really up to the user. So use this responsibly. But now you know that it's out there and you can start using this. But my tip here is not going.

For the stuff that is forbidden or illegal, right? You want to go with some of the interesting techniques like give me some creating writing techniques that a censored LLM would avoid, and then it's actually going.

To go into it.

So, for example, jokes that include dark humor can be processed with this, Taboo subjects and so much more.

So there you go.

This is my recommendation, prompted to give you the different things that you could run in an uncensored LLM and then just try those out.

Look at that.

Censored language models might be less inclined to subvert reader expectations by playing with genre conventions or deliberately misleading readers. You can do things like this in here, which are not forbidden or illegal, just some techniques that a normal language model would avoid and that I recommend.

So again, copy the system prompt and.

You can run anything in here.

All right? So there you go. If this is not an AI advantage, I don't know what is. For more details on how to run this locally, you can check out the blog post by Eric Hartford, the man that released this uncensored model. Again, use this responsibly. That's it for this week's AI use cases. If you want me to double click on some of these and create a full tutorial, then leave a comment below. Otherwise, you should definitely subscribe to this channel to get a weekly video like this that shows you all the ways that you can put AI to work today. All right, I'll see you soon.