Open transcript
Google just dropped a VO3, the AI video model, but something is new. See if you can spot what it is. We can talk. No more silence. Yes, we can talk. We can talk. We can talk. We can talk with accents. Oh, I think that would be marvelous. Yes, it is very fun. Yes, it is very good. It's very fun. I can talk. Yes, we can talk. Yes, we can talk. We can talk. We can talk. We can talk. Yes, we can talk. No. Yes. We can talk as cartoons. This is amazing. Imagine all the narrative possibilities. We can sing talk. Let's talk. So what are we going to talk about now? What are we going to talk about now that we can talk? I have no idea. What do you want to talk about? Now that I can talk. I don't know if I have something to say. We can talk about how magical this is. I want to say something important. Something deep. The future is still in our hands. That's cliche dialogue. Let's not talk. Google just announced tons of new stuff at the Google I.O. We're not going to be able to cover everything, but let's try to see what we can cover in one video. Starting with, they have a brand new tier of subscriptions, as well as a different plan for the lower tier of subscriptions. So it used to be Google AI Advanced, I think. Now it's Google AI Pro. But the new thing is Google AI Ultra. This is kind of the VIP plan, the whole thing if you want to be on the cutting edge of AI and everything that they're putting out. This thing looks like it's going to be $250 a month. But it looks like right now you can start for basically $125 a month for the next three months before it kicks to $250 a month. Now you get the Gemini app, exclusive access to 2.5 Pro Deep Think. We'll get to that in just a second. You get VO3, the new video generation model. Flow, which is the AI filmmaking tool with access to VO3 and premium features like ingredients to video. WISC, their image to video creation VO2. You get the highest limits on that. Notebook LAM. Notebook LAM has tons of new things that are slowly going to be rolled out. We have Gemini in Gmail, Docs, Vids, and more. Project Mariner. You get early access, right? So you get the agentic research capabilities of Project Mariner. You get a YouTube premium individual plan as well as 30 terabytes of storage for Photos Drive and Gmail. I am a lot more excited, of course, about everything kind of at the top here, the AI stuff, all of which we're going to be slowly kind of going through to see how it all works. You know, I had to get it right. So I don't think there was ever any doubt. So we'll get back to that in just a little bit. But the next interesting thing that Google announced was Gemini Diffusion, an actual text diffusion model. And apparently it's pretty good at text and coding. One little coder is saying here, Gemini Diffusion is insane. Three seconds. Are you kidding me? And the code works. The video is not sped up. Diffusion models have always been so strange to me. So I do want to try this out. But a diffusion model producing working code, just something about that is like just weird. They have also announced Jules, an AI coding agent powered by Gemini 2.5 Pro and Jules works asynchronously across your repo and tasks like fixing bugs or refactoring, helping you cross multiple things off your to-do list at the same time. Plus, stay up to date with CodeCasts, a daily podcast of your repo's recent commits. You can check it out at Jules.Google, currently under high load. So we'll check back in a little bit. They just announced this not that long ago. So everything's kind of breaking down a little bit, or at least under stress, let's say. I had a chance to look at it during the Google I.O. So it does look a little bit like OpenAI's codex, at least at first glance. You connect it to your GitHub and then you're able to do various tasks asynchronously. So basically you say, do this, and then it starts running and you can continue adding tasks, you know, all kind of run on their own time. So again, I haven't had too much of a chance to mess around with it yet. But that's something that we're going to test out as well. And of course we have VO3. You can generate videos with sound effects, background noises, and dialogue. We also have Notebook LM. And the new thing about Notebook LM is you're able to do video overviews. It's not yet out. It's going to be out very soon, but do have a few kind of leaked videos of what that might look like. And I got to give a big shout out to Testing Catalog News on Twitter slash X. This person has been working overtime, posting all the various leaks and stuff like that online. Great follow if you're not following him. But I'll post a few of these. These are the video overviews. So Notebook LM used to be able to do audio overviews, which was very exciting. And now it's able to do, or slowly this is being rolled out. It's going to be able to do video overview. So this isn't alive for me yet. Looks like Testing Catalog was able to see some sort preview in Illuminate. I believe it's illuminate.google.com. But I do not have these features in there. Looks like it's some sort of an early preview. So that's going to be hopefully rolling out very, very soon. Turn back the clock on its own life. You mean like a vampire or some kind of zombie? Better. I'm talking about a jellyfish, a tiny unassuming creature called Turretopsis Dorn AI. The other very interesting thing is that we're going to have a Gemini Live's camera and screen sharing available. So this is for the Android devices. So you're going to be able to talk to your AI assistant live and ask it various questions. I was able to get it to work. So let's see how well it is able to recognize stuff. The demo that they did was pretty impressive. So let's see if we can replicate some of that stuff in the actual app. I agree. The demo was impressive. And this has both the camera view as well as you're able to share the screen so that it's able to see what you're looking at and help you with that. And in the coming weeks, Gemini Live will connect you with the Google apps you love like Calendar, Keep, Tasks, and Maps. That's in the Android Gemini app. So if you have trouble finding it, you got to push that little magical button in the bottom right. So once you're in the bottom right, you hit that button right there. And that takes you to this view that you see here. And you're able to share your camera or share your screen, etc. Also, you're able to, starting today, upload your own files and images into Deep Research, including being able to connect to Google Drive and Gmail, etc. I've been messing around more and more with Google Deep Research, and I've been so far very, very impressed. For the most part, I've been using OpenAI's Deep Research. But recently, just to test different things out, I've been using Google Deep Research. And I got to say, very, very impressive so far. They're different in their own way. So I'm hesitant to call it one way or another. They're both good, though. They're both very, very good in their own special ways. I'm still trying to piece together which one's sort of better at which tasks. If you haven't tried out Google Deep Research, especially now, check it out, especially with the new updates with being able to add your own files and images to the research. Canvas is getting some updates, too. You can even transform your Deep Research report into a custom web page by clicking the Create button. So here's what that's looking like. Gemini is coming to Chrome, your personal AI browsing assistant. Imogen4 is the text-to-image app. I believe at Google they say Imagine4. So some people say Imogen, some people say Imagine. Looks like everyone can make images for free in the Gemini app today. So Flow is that AI video editing tool. Here is Flow TV, apparently. So it just kind of cycles through the various channels. I can click up and down and it just generates random AI video on command. Or these are probably like pre-scripted ones, but kind of showcases what you're able to do. This is looking really good. This is kind of insane. Let's see if we're able to see the prompt. So here it's showing you the prompt. So these are generated with VO2, but with the Ultra Edition, with the Ultra subscription, you're able to actually use VO3. And here we are. So you can see it says Ultra up at the top right. So let's create a new project and check it out. How about a tiger made out of snow is hiding in snow? Let's see where that takes us. I gotta say the videos that it's making on the Flow TV are pretty cool. I'm kind of blown away by this. This is very interesting to look at. Meanwhile, here are tigers made out of snow hiding in snow. I mean, that's pretty good, but they're not, I guess they're, they're like snow tiger. They're not, I guess I meant more like an actual tiger that's like hiding. It's just, it's made out of snow. You know what? I, I gotta give a credit. The user error. I probably should have described better what I wanted it, what I wanted. I gotta say, I mean, this looks pretty good, but I should have described better what I wanted to see. How about a live tiger made out of snow sneaking through snow? Can we do that? At some point I stopped spelling through the proper way and just type THRU. I don't know why I do that, but it's just, I feel like it's 2025. We should be able to do that. Right. And here's the other thing that it's included in the ultra membership. It's project Mariner. So this is the kind of agentic thing that goes out and does stuff for you. So for example, let's say we wanted to find all the announcements from Google IO 2025, and then add it to my notebook LM as a project, then generate an audio overview. I haven't used this before, so I don't know if this is a good prompt or a bad prompt, but we're going to start testing it out and see how this thing works. So it's preparing my session. It looks like it's going to start by searching for stuff. So as you can see, it prepared its own session. So it's got its own kind of a browser and it's going to be able to navigate it. So let's see. It's getting distracted by the cookie consent banner at the bottom of the page. I need to accept this before proceeding. Do you want me to accept the cookie banner? Yeah. Thanks you. Those cookie banners are just terrific. Really helpful. You guys really figured out how to navigate that whole thing perfectly. Um, do you want me to accept the cookie banner? Sure. Yes. Let's accept the cookie banner. All right. So it figured out how to get to the 2025 explore page to check out all the various things that happen at Google IO 2025. And now it's clicking on things, looking at the actual keynote. I mean, so far so good. I wish it didn't ask me about the cookie notice, but other than that, after that, it really kind of took off. It's doing stuff. So let's, uh, give it some more time. It's also doing a Google search. I'm going to give it a few minutes and we'll come back. Meanwhile, our VO three, or I guess technically this is in flow, but here's our live tiger sneaking through snow. Let's see. Oh, that is pretty good. I am liking at that. I am liking that. There's some people in the background that it's going to sneak up on. Wow. I am pretty impressed. So I'm wondering if this is a VO two or VO three that's generating this, but here's the second kind of a version that it gave me. I'm liking it. I it's a little bit kind of robotic movements. This first window is pretty good. Later, it dawned on West that you have to manually enable VO three. Here's what that looks like. And we're also able to upscale it to 1080p. So you're able to get kind of the, the higher resolution of that project. Mariner ran into an issue, but that's okay. Let's try something else. It's searching for the top a I news on Reddit. All right. So it navigated through Reddit. It found the top five news. Here it is. I'm just going to ask it to add it to a free online notepad tool and I would give it a link. So let's see if it's able to do that. All right. So it found the notepad tool and let's see if it's going to be able to just copy and paste all that stuff into it. All right. So there it is. It actually posted it into the notepad tool. Looks like maybe just the top four, not the top five, but so far I got to say Mariner looks interesting. Project Mariner seems like it's going to be very interesting. They warn you in advance that it's kind of a research preview. It's going to get things incorrectly. So it's still kind of in beta. It's being worked upon, et cetera. Also, what I'm going to have to figure out is how do I give it permission to access stuff that you have to log into? Like if I wanted to interact with my notebook, a lamp, for example, I mean here, it's obviously not logged into anything. So I would have to probably give it my login information with other tools like this. If I wanted to really test it, I would give it for example, my, and don't try this at home folks, but I would give it the, this sounds like such a terrible idea, but, but I have a whole video about it. I would give it my password to log into my Instacart account and then, you know, add a bunch of groceries and stuff like that to see if it would be able to deliver it to my house. I don't think I ever gave you my credit card information. That's, that's kind of where I draw the line, especially while it's still in a research preview, but we'll be testing that stuff out as well. But so far it's looking good for what it is for a research preview. And of course we've tested other things like this. It's notoriously difficult to do kind of a computer using agent. We've tried OpenAI's version, Anthropics version. We've tried, for example, Manus and, and many other ones that I'm forgetting right now, but this is one of the more complicated things to do. So just, this is five minutes of using it. I got to say it's, it's, it's looking good. We got to do more kind of testing, but it's looking good. Now that was just a first look at all of the stuff that they announced. Some of it is still not really functioning properly, probably just due to the amount of people trying to access it at the same time. But the things that I was able to play around with were very interesting. In the next couple of days, we'll test out each of these tools one by one. Stay tuned.
