This Week's Unique New Ways to Use AI

This week, we got a wide variety of AI use cases spanning across the leading voice generator, introducing a brand new suite of tools to translate, for example, a YouTube link into another language. Then we have some surprisingly good TikTok research and a completely free tool that turns anything shot into a website. I'll show you the entire 20 second workflow. It's weird and better than GPT-4 and more. So without further ado, let's dive into all the AI use cases that came out over the course of the last week. And as per usual, we'll be focusing on the ones that you can put to work today.

So let's start talking about 11labs dubbing studio because as a content creator, but also a person that moved abroad recently, this is extremely exciting to me. Translation is getting better and better. And in my humble opinion, 11labs was the leader in the market when it comes to quality of this AI generated voices. And now on their website, they added this brand new tab, which is called dubbing. The thing is, this is very simple to use. So I'm just going to go from English to let's say, Spanish. And as you can see, you can share various links here. So I'm just going to take the shortest YouTube video I have as this is limited to 10,000 characters with the free plan. And there you go. It's dubbing it into Spanish. In other words, it's not just doing the translation, it's also creating the voice in the same tonality that I speak in. Let's see how this goes. So that's not bad. And the cool thing is that YouTube is introducing this feature natively into the interface, as you can see here in the bottom. So this is gradually rolling out. I hope everybody will have it soon. The AI Advantage channel doesn't have this yet. So as you can see, there's a little switcher here at the bottom where you can actually switch the audio tracks. These aren't just subtitles, but these are fully dubbed tracks, like movies dubbed into foreign languages. What this means is that very soon all video content on the internet should be available in all languages. That's pretty exciting, especially if you consider other apps that actually adjust the video to fit this. Like for example, Heygen. I just quickly wanted to point your attention towards this because last week there was a speech by the president of Argentina that you might have caught and the entire thing was dubbed into English from Spanish. But they use this tool called Heygen that also adjusts the mouth moving. So in my opinion, this is way better than the original with the dubbed voice over it. Today I am present to inform you that the Western world is facing a significant threat. Tech like this is sure to change the internet in 2024. But now let's move on to the next use case that you could be putting to work today.

And as we're talking about combining multiple AI tools, this is one of the biggest points I would like to communicate in this format, right? We look at the newest use cases that you can put to work. But this doesn't mean that these use cases should be used in isolation only. These usually stack beautifully on top of each other. All you got to do is get a little creative. And here's some inspiration for that. So this is a combo of some tools that I showed you over the last week and the results are quite stunning. So look, the workflow starts out in Craya where you can paint and it generates the image with stable diffusion. And then this image that is generated, he runs through Magnific to increase the sharpness. Look at that. Before, after, that super sharp custom image he puts into Magerni to remix it with another prompt, which will generate the most artistic styles as Magerni is the best at that. Then he takes that, just adds a few random scribbles in Photoshop, uses Magnific that we talked about to scale up the entire thing. And look at this. Isn't this incredible? So shout out Emily Art for doing this. I love to see these combinations. If you come up with something similar, feel free to send it in via email. But now let's move on to the next tool. And next up, we have Runway ML with a new tool that allows you to use multiple motion

brushes to animate photos into videos. So if you're not familiar with the original feature, it's very simple. You just paint over a certain area and then you define how you want it to move and it just does it. The problem is that in most scenes, you want different types of movements, right? The clouds move differently than the foreground subject. Now you can do that and these are the results from the trailer. Multi-motion brush is bringing a new level of control to Gen 2 and we're really excited to see the creations that you make with it. So yeah, humans still look weird, but I found this fantastic little example on Twitter with somebody using an anime scene. So as we talked about, usually these more anime, cartoonish, or drawn styles work better because they're more forgiving. Hyperrealism is really the hardest thing here.

You might've heard that in today's digital landscape. Data is the new oil. The success of your online business often hinges on the quality of the data you collect. But here's the challenge. Obtaining high-quality, relevant data is easier said than done. Enter Bright Data, today's sponsor that helps you turn these challenges into opportunities. So let's take a real-world example here. Imagine an e-commerce store struggling to understand customer preferences leading to poor sales. Or consider a travel agency unable to track real-time flight price changes, thereby losing its competitive edge. These are just a couple of scenarios where the absence of reliable data can be a major setback. But luckily, Bright Data specializes in collecting and transforming raw, unfiltered web data into valuable, structured insights. How do they do it? For a technique known as data scraping, which is essentially a bot going around the internet and collecting all the data. That's also how all of these large language models get their training data. But data scraping comes with its hurdles. Rapid and large-scale retrieval can often trigger websites' defenses, blocking your access to it. And this is where Bright Data stands out. The robust proxy infrastructure and web unlocker tool ensures seamless access to the public web without interruptions. And additionally, Bright Data also offers a dataset marketplace that is full of ready-to-use datasets from social media and e-commerce datasets or data that is used to train large language models. And if you have specific needs, their custom dataset service is there to cater to your unique requirements. So to sum it all up, Bright Data equips you with the tools to not only access essential data, but also to transform it into actionable insights, which in turn allows you to power your business strategy and stay ahead in your industry. If you're ready to turn information into your business' most valuable asset, then click the link in the description to get started with Bright Data today. And now let's get back to the next use case that came out over the course of the last

week. And this one's actually less of a tool and more of a feature. So Google Chrome added this new AI image generation to customize your Google Chrome. Now I updated mine, I don't have it yet, but I think this is a fantastic example of where all these products are heading. As I usually say, first we get the standalone tools, and then they just blend into the applications we're already using, as this is the next evolution of software. And this just illustrates that so beautifully, right? They have an image generation model, and now they just integrated it into Chrome, which allows you to customize your experience a bit. This is another thing I would expect from all the tools we're looking at, right? They're just going to be integrated into the apps.

And one obvious one here would be TikTok, right? And this is a bit of a controversial topic, as there's a lot of privacy issues surrounding the company, and the app gets a lot of people fired up. But the fact is, they did become one of the biggest video platforms on planet Earth, and they have the data to do things that nobody else will be able to do. So this is them acting. And look, this one particularly might not be something you'll be using today. Okay, you could, and I'll also show you how. But the point is, this is something that you're only going to be able to do if you have a data set like TikTok does. So what this tool does is create so called depth maps from video. Depth maps are used in 3D modeling to communicate depth inside of a scene, okay? So the different tones that you see on screen right now represent different depths. So obviously, bright yellow is closest to the camera, and black is furthest away with various tones in between. And obviously, in the demo, this looks really good, they seem to get it right, and they're able to separate the background from the subject and all layers in between. Now, sure, you could go ahead and create your own depth maps if you're into 3D modeling. Right now, the demo space only works with images, their demo is with video, but that's fine. So you just upload an image here, we'll just take one of the presets, and it generates a depth map like this. So I don't know, I realize most people won't be doing this, but I just wanted to share this as I thought this was super unique. And if you want to do stuff like this, now you know what to explore. A good starting point would be probably downloading Comfy UI for the Pinocchio one click installer, and then downloading some of these notes and workflows. And this concrete one you have in this thread that I will link below. And with that being said, let's talk about the next use case, which is not live yet.

So I was hesitant about putting this one in, I just think it's too interesting not to. So you could make this work for GitHub today. And also there's this demo space on model scope. And let me tell you, I spent about 20 minutes trying to figure out how to log in here with different translation tools. I speak multiple languages, not this one, I wasn't able to log in. If you have some tips for me on how to do this, please share them in the comments below. Basically, here you could use this. But what this motion shop does is replace video characters with 3D avatars. So essentially, it cuts out the character, it reproduces the background, and then maps the motion onto a 3D character. This is going to be huge for AI and VR applications once the technology is really there for this to be done in real time. So yeah, this is just an interesting one that I wanted you to see. Some of this video research that is coming out is becoming increasingly mind blowing. But now let's move on to one that you can actually use today, because that's the entire point of this series.

And this one is the most popular hugging face model of the week, called screenshot to HTML. And at this point, you might be saying, Igor, but you can actually use GPT-4 to upload a screenshot and turn that into HTML. Plus, there's some other tools. Why should I care about this? Well, fair enough. GPT-4 does that, yes, but it's paid. So our most other tools or things like GPT engineer are not freely available yet. And this one is, and it actually works. So let's just use this and compare it to GPT-4. So what I'm just going to do is pull a screenshot from the Apple website, okay? Then I'm going to upload the image to this tool, also simple. And once I click submit, it generates the HTML. Now for comparison's sake, I did the same thing inside of GPT-4. I just use a simple prompt that says write the HTML for this. And GPT-4 did just that. Nice. So it did note that I need an extra CSS document. So let's just copy this and go to one of these websites that visualizes HTML code for you. Okay, so literally just paste this in here. And this is what GPT-4 created. Okay, so this is very bare bones, nothing close to the original. But to be fair, you would need some CSS to style it the way it's on the website. And the cool thing about screenshot to HTML is that actually includes the styling. All right, and after two failed tries, as a lot of people are using this right now, here's the HTML for us to copy. So look, all I do is just click copy. And in the viewer, I paste. And there it is. Look at that. This is really close. So sure, I would have to replace the images. But basically, this got us really close. And to be fair, the GPT generation did not include the CSS, which is there to style the website. Okay, so that's why it looked so bare bones. But then I went ahead and I told GPT that, hey, I got this code from a friend of mine, you know, this being the friend, I didn't want to insult the AI with the fact that I'm actually using a different app. Anyway, I gave him the code. And I asked what is different in this. And it told me about all the differences. And then I told it to rewrite the code to be just like the code from my friend. And even after that, the result is not as what we got from screenshot to HTML. So completely free, useful, and even beginners can build websites like this from a screenshot. Now, remember that you can always use GPT-4 to help you along the way if you have absolutely no idea what you're doing.

And there we go, this week's use cases. If you care to learn more about what you could use AI for today, I do one of these episodes every Friday. And other than that, we do a weekly newsletter that is completely free. And when you sign up, you get an ebook and a notion template with prompts, styles, GPTs, prompt generators, personal prompt database, all for free. So I hope this was helpful, and I'll see you next week.