Mind-Blowing AI Tools You Need to See

Matt Wolfe demos AI tools for creating games, 3D scenes, avatars, and music from text. FR VR generates games from prompts, GPT Engineer builds apps by asking questions, and Infinijn creates procedural 3D environments. Music Gen makes 30-second tracks, while QR art generators turn codes into images.
Questions this video answers
June 21, 2023
frvr.com/ai
Over 24,000 stars (and over 5,000 on the day of recording)
At least 16 gigabytes of RAM
About 2,361 stars
Chapters
So I just got back from a camping trip, had to go out and touch some real grass for once. But of course I can't go anywhere without taking some tech with me and played around with my drone. But I wanted to get back and make a video about all of the cool tech that I've come across over the last couple weeks that maybe you haven't seen yet that you probably want to see. I recently came across this demo for a product by a company called FR VR. And it's a company that claims that you can do a text prompt to another example here, top down 2D space shooter. And this just writes out the entire code for the game and they have a playable space shooter. They modify some of the stuff with the bullets here and you can see it changed the bullets. They had a prompt at the bottom. Bullets do not seem to do anything when hitting enemies and it fixes that enemy should fly into somewhere in the top half of the screen after spawning and then it shows them moving around. Integrate a scrollable star field and you can see it added stars in the background. They just simply add these text prompts into game and it tweaks the code and fixes the game for them. Then they go and they add some new image directions. It creates the spaceships, it creates alien enemies. And as you see, it's building this whole game in real time. It generated the assets with AI and now you have a spaceship shooting alien spider creatures. And this is their final product and it's a company called FR VR. I don't know if that's pronounced forever or fervor or if it's just frvr, but they have a wait list that you can get on@frvr.comai what's interesting about this is I made a tweet not too long ago about how in a few years you'll be able to do text to game where you just type a prompt of a game you want to see and it will just go and create that whole game. I said we'll probably see that in a few years. Literally within days of writing that tweet, I come across this. Now this is cool and all, but I don't know when we're going to see this. I don't know when it's going to be available. Right now we just have a wait list. This could still be months off before anybody ever sees it. However, I was scrolling the trending repositories on GitHub as I sometimes do, and I noticed that the number, number one trending repository is this Thing called GPT Engineer. Specify what you want it to build, the AI asks for clarification and then builds it. It's got over 24,000 stars and over 5,000 stars on the day I'm recording this video. This, to me reminds me of all of the excitement that was going around when Auto GPT came out. This is like that, but it's a GPT that will build any app or game that you can imagine. And this is available right now. You can install it on your local computer and run it. It's open. I'll link below to the GitHub collab because there are installation instructions. This is probably something that I'll do another setup video in the future and actually walk through setting this up and try to get it to build some stuff for us and see what it can do. But it doesn't look too difficult to install on your own computer and run yourself. Now, here's the demo they give. It's not quite as visually appealing as that FR VR demo we just looked at. This is what it looks like. It kind of looks like what Auto GPT looked like. You're just in a terminal like environment here. But they type in Multiplayer Snake in the browser. Use a Python backend with MVC components. The view needs to stream the state to all connected players. Please implement also the HTML and JavaScript necessary to run the game with only the code you generate. And then it asks for some clarification here. So it says game rules and mechanics. How exactly does the Snake move, grow and interact with other players? Player connections. How many players can join a game? Is there a lobby system or matchmaking? Game state updates? How frequently should the game state be updated and streamed to the players? What is the desired latency? User interface. What should the game interface look like? Game controls. How do players control their Snake game end conditions? How does the game end? Is there a scoring system or leaderboard code structure? Are there any specific requirements for the organization of the Python backend? And they just answer all the questions and then it just goes off and writes the code. You can see it's just generating the code for a Snake game. Create some HTML css. They take this code and it looks like automatically just added into Visual Studio and you can go and run that code. And this is the simple Snake game that it generated. It eats the little red dot and then the Snake grows longer. It eats that red dot. It goes longer. I mean, it's a Snake game. What you'd expect from the game Snake. And here's another example that I came across from Anton, who's actually the creator of the GitHub repository we were just looking at. And in this one he's creating a Pomodoro timer. You can see his prompt. Develop a Pomodoro Timer app using HTML, CSS and JavaScript. Allow users to set work and break intervals and receive notifications when it's time to switch. The app says, we will create a simple Pomodoro Timer app using HTML, CSS and JavaScript. The app will have the following files. It lists off the files. It's going to have Index HTML style CSS script JS package JSON let's start with the entry point file. Index HTML it goes encodes the index HTML file. Now let's create the CSS file. It goes and builds up the CSS file. Now let's create the JavaScript. Goes and builds the JavaScript and then now the JSON file, it builds that and then it says press Enter to run it. And then it opens it up in a browser and they've got a working Pomodoro timer. They can set the time in minutes on the left and the amount of break on the right and they can press Start Timer again. This reminds me so much of what everybody was so excited about, Auto GPT. This actually feels so much more useful than what Auto GPT was actually able to do for people. You prompt an app that you want, it will ask you questions to clarify what's going to happen with that app. And then once all of its clarifying questions are answered, it goes and just builds that app for you. Now I'm sure it's a long way off from building the next, you know, Zelda, Tears of the Kingdom or Elden Ring type game. But for real basic games or very MVP low level games, I mean, you can go and build something really basic right now with this. And then when that FR VR tool comes out, it looks like you'll be able to take it to even another level. Now here's something else I came across recently that just absolutely blew my mind when I came across it. It's called Infinijn. And this is actually not using AI. This is procedurally generated 3D scenes and it's totally free and open source. And it uses Blender, which is a 3D modeling program, which is also free to use. And let's just check out this demo here. You can see this 3D scene that's appearing right now. Was Gener using this Infini gin tool. These fish swimming through the water was generated with this tool, this snowy landscape generated with this tool, I mean, these are so realistic. And this is all completely free to use in Blender. Now, again, this isn't AI. This is procedurally generated, meaning that it kind of follows a set of rules every single time. But, I mean, the landscapes and the realism that we're getting out of this video just looks amazing. And you can get as detailed as you like because you have control over this inside of Blender, you can see all the various layers that it adds on top of each scene. And it just gets more and more intricate as it creates all of these things. The data is labeled so it knows everything that's there, so that you can actually manipulate and edit every element of every scene to the most granular detail that you can imagine. You can create unlimited variations, randomized, automatic and controllable. So you can see it's changing the seed number here to get different scenes. Scenes changing the amount of cloud coverage, how much the water height is, the tree height, the bark random seed, the leaf random seed, the twig density. Everything is controllable to the most granular detail with this application. And then you've got this, like, iguana here. And you can see it's changing the head size and the tail size and the leg size. And of course, like I mentioned, it's free and open source. You can literally create anything you imagine with this. And so I'm just thinking, for game worlds and for 3D movies, this is going to be a game changer. Now, this is also another open source tool that is available right now on GitHub. In fact, if we look back at our trending GitHubs today, this is actually the third most trending today with already 2,361 stars. Now, in order to run it locally, it does require at least 16 gigabytes of RAM. And what I understand from other people who have already played around and tested it, it still loads pretty slow. So expect a long time to generate the scene that you want to generate. But honestly, this is really just showing you the pace at which this technology is moving and how quickly any individual will be able to create 3D worlds for video games or movies or whatever it is you want to create. It's becoming easier and easier for just the normal person who doesn't know how to do crazy graphic design to do this kind of stuff. All right, so I'm here putting some finishing touches on my edit. It's actually the next day and I'm recording this because I came across this tool that I Felt had to make this video. It's called Instaverse from Illumina. And you can play with this right now. And it's free to play with. At the moment, it's at Illuminai, GitHub, IO instaverse. I'll make sure it's linked up below. But check this out. I can go over here. It uses Blockade Labs, the tool that makes like a 3D environment. And we can create any 3D world we can imagine. Let's use a surreal style here. And let's do inside a futuristic space station with windows that look upon the universe. And here's the world that it generated for us. I can actually drop a character. Boop. Right into this world and move this character around in 3D. And it actually uses the depth map of this graphic to know where the character is. Make the character sort of go uphill downhill and run around in this 3D scene that we just created. You could also view some of the other scenes that other people have created on here recently. Let's try this one here. All right. It dropped my charact character into this scene here that this person's created. And I can make my character run around in this creation that somebody made. Here's another one of this outdoor environment. Some of the aspect ratios are a little funky. That needs some work. But this is just the beginning. This is as bad as it's going to get. Notice how when I run up against this rock right here, he kind of gets stuck. And if I go around, I can actually walk up on top of this thing here. And then he kind of drops down again. Still super early, but super exciting to see where this stuff is going to go. I just thought this was cool enough to change my clothes and add a quick recording into this video to show off this tool as well. So here's another piece of research that I came across recently. It still says the codes to be released, but it's called Avatar Booth. High quality and customizable 3D human avatar generation. This is basically a text to Avatar generator. You can see here's Obama, Hillary Clinton, Kobe Bryant. Now, I don't think this is going to win any awards for best character avatars. I mean, they kind of look lumpy, and I don't know what's going on with their legs. They all have kind of way longer torsos that they have legs. But from a technology standpoint, I mean, we're getting to a point where we could literally just type a text prompt and get a character. I mean, this is only going to get better. Like we've said many, many times before, this is the worst it's going to get. Which is great because I think some of these might give me nightmares. Here's some more examples with Bill Gates here, Abraham Lincoln, Kobe Bryant just looks hilarious to me. I kind of want to 3D print that just to see what that. Yeah, I just want that on my desk. I will say this Eragon looks pretty good. I think it's because he's wearing this long, like cloak thing. It sort of hides the fact that probably has stubby legs just like everybody else. The stormtrooper doesn't look too bad either. It's also going to have character personalization where you can upload a picture of your face and put your face on the various 3D models. You can upload a picture of clothing and then put that specific clothing on the characters. Somebody probably stuck them in mixamo or something and animated them. I don't know, they're kind of funky looking, but I'm kind of looking forward to playing around with them and making some characters out of it. But I just got to remind you, this is the worst is going to get. All right, so last week I talked about Music Gen, which was Meta's new platform that can generate generate songs from any prompt. Well, this is a tool that just became available over on Replicate, where you can generate up to 30 second songs. So if you remember the Music Gen hugging face space, it would generate roughly 10 second songs. Bluegrass, banjo, heavy metal. And here's what we got. You can see It's a full 32nd song now. Banjo solo. So we can get a little bit longer music now with this waveformer. And again, I'll make sure this is linked up below so you can try it if you want. It's free to use and it's available on the Replicate site. All right. In last Friday's news breakdown, I talked about the hype around the QR craze that's going around where people are taking QR codes, running them through Control Net and creating cool images. And sometimes they're scannable, but I found that most time they're not. Well, it didn't take very long and people have already started to turn this into apps and easier ways of doing it. One that I came across is this Qrcraft X xyz. You give it a prompt here. A painting of a Japanese house in the snow with a sky background. Andreas Roca Matte painting Concept art. A detailed matte painting. Future Tools IO Creativity Strength. Let's leave it at one. I'm just going to leave all these at the default. See how they do it says here that this QR code influence. Wait if the QR code doesn't scan to up this value and maybe it'll scan better. Let's go ahead and generate qr. It says I have one generations remaining, so now I have zero. And of course, if you want more than one generation, you got to go pay for more generations. It also says please wait one to three minutes, but that literally took less than a minute. And here's our generation. Let me see if this scans when I try it on my phone here, right? Not scanning. Sometimes I find if my phone's too close to the computer, I just back off a little bit and I get further away. It scans. It's finding the QR code, but it's actually not going to the URL. You can try it with your phone. Maybe it'll work on an Android, but it's not working on my iPhone when I try to scan this one. I don't have any more generations, so I can't test QR craft to see if I could get one to work the next time. Oh well. What I'll do instead is move on to this free hugging face space that I came across that does the same thing, but for free. So this is a QR code AI art generator on hugging face. I will make sure that this is linked up below so you can use this one as well. And it literally does the exact same thing. And let's see, this is their default one that's here without me actually adding a prompt. Let's see what happens if I try to scan this one. So this one actually scans. This one takes me to hugging face. Let's go ahead and try to create our own on this hugging face space. Let's try future tools IO. You could upload your own QR code if you've already got one, or it will generate one for you. Let's do a family of wolves in a snowy mountain scene. And then I'm going to leave all these settings the same. I'll set the seed to negative one to make it random. And then I believe this control net conditioning scale. It's sort of the same thing as this QR code influence weight. So if it doesn't scan the first time, I can up this conditioning scale and and get the code that I'm looking for. So let's go ahead and run this one. That is going to take about a minute and a half. And here's the QR Code that it generated with a family of wolves in the background. Let's see if it actually scans. But this distance, it doesn't scan. But I'm going to scoot back a little bit, because sometimes it will. Ooh. And it found it. If I click on this. Takes me to future tools. So this actually worked. This QR code worked the first try. Hell, yeah. That's pretty cool. So this is QR Code AI Art Generator. It's a space on hugging face. And speaking of QR code, this has to be the coolest QR code I've come across so far. This is from my buddy, Ian Curtis. He does some really, really cool stuff in the extended reality space, mostly with augmented reality. He created this QR code, and watch what happens when I scan this QR code. So we'll scan it here. And, oh, my gosh, the QR code's falling apart, and there's an alien popping out of the screen, and it looks really kind of crazy. And I can actually look at it from different angles and look around it and zoom in on it. And look at that. You can see it's like this moonscape with Earth in the background. And I can actually, like, move my phone around, and you can see it at different angles based on where my phone is looking at my computer screen. Like, how freaking cool is that? I have no idea how Ian made this. I'm assuming it's using this 8th wall tool that you can see in the little URL bar at the bottom of my phone here. But this is something I'm gonna have to learn more about because I want to create some cool stuff like this, where you scan a QR code on Twitter, and it creates some really, really cool effects. I'm showing you this in real time right now on my phone. And finally, I want to show off a real simple but really useful tool from. From my buddy Chris Haydorn, who, if you're not familiar with him, he has a killer YouTube channel over at Tokenized AI, where he really, really hones in on some of the best Mid Journey tutorials and tactics. But he created this Mid Journey stats tool where you can find out how busy Midjourney is at any given time. Mid Journey might load slower and faster, depending on what part of the world that you in. I'm in the Pacific time zone, and right now, if I wanted to use Relax mode, it would take me about 84 seconds to generate an image with version 5, 570 seconds to generate AI art with NII 5, or almost no time at all to generate with version 4 you could see different times of the day. If I'm using fast Mode, maybe about 8:30 would be the worst time to be generating. But the rest of the day, you know, fast mode's gonna be pretty fast. I should probably avoid generating in the middle of the night because this seems to be when the load times are the slowest. And again it different based on the time zone. If I go to Hong Kong and apply here, you can see right now if you're trying to generate with version 5 in Hong Kong, it's going to be pretty dang fast. In fact, it's going to be pretty dang fast no matter which mode you generate in. Maybe you're generating an image with mid journey and it's loading really slow. You can find out, is my computer messing up? Is it generating or is midjourney just slow? Right now this is a really handy tool. You can find it over@midjourneystats.com and again, I'll link this up in the description. And finally, I'll wrap up this video with something that Jensen Huang, our GPU overlord, recently said at the Cannes Lions Festival. He said for the very first time, the creative process can be amplified in content generation. And the content generation could be in any modality. It could be text, images, 3D videos. These tools aren't a replacement for human creativity. They augment the skills of artists and marketing professionals to help them feed demand from clients. By producing content more quickly and in multiple forms tailored to different audiences, we will democratize content generation, Huang said. And I just love these statements because there's so much fear right now that AI is taking jobs, and to some degree, some jobs will become obsolete because of AI. But creative professions. I still feel to this day that using AI amplifies creators. It makes the lives of creators easier. As Huang says, it supercharges them. If you're not very creative and you don't have very many artistic skills, now you've got some because of AI. If you are creative and you do have artistic skills now, you've got frickin superpowers to add onto it. And that's what I love about AI. And that's why I love this message from Jensen so much, is because it sort of echoes my thoughts about creativity and AI. I think the artists that are embracing AI are seeing that, holy crap, this actually takes my art to a whole new level. And those that aren't artists that are embracing AI are starting to see, wow, I can actually do things that I wasn't able to previously do, which is why I personally am so passionate and excited about AI. And if you get as excited about this stuff and you want to stay in the loop as much as possible, check out FutureTools IE. This is where I curate all the cool tools that I come across. I'm adding new tools every day. I also keep this site up to date with all the latest news every single day. And I have a free newsletter where I send it every Friday and it's just the TLDR of the Week. I'll send you just the five coolest tools that I came across, a handful of news articles, handful of YouTube videos, and one cool way to make money with AI. It's the TLDR, the grand overview of the week in AI. It's read by over 115,000 people to keep their finger on the pulse of AI in just one weekly, easy to read email. You can learn more about it at futuretools IO. Thank you so much for tuning in. I really appreciate you. Thanks for nerding out with me. See you guys in the next one. Bye Bye.










