Everyday AI Podcast – An AI and ChatGPT Podcast - EP 619: Nano Banana Uncovered: A practical guide from inside Google with Paige Bailey

Starting point is 00:00:00 This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the all-in-one creative AI studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. If you don't know anything about nanobanana, seriously, where have you been?

Starting point is 00:00:50 Let me drop a number here. Five billion. That's how many images have been made inside the Gemini app with nanobanana in less than a month. And maybe you're thinking, oh, okay, Jordan, like, you know, I'm not like a creative person. I'm not a technical person. And why would I use nanobanana? Well, you're going to want to listen to today's show and stick around. Trust me, because we're going to be uncovering some of these secrets and real power behind nanobanana

Starting point is 00:01:20 and how you can even, in natural language, create an app in a couple of minutes that is actually using nano banana. It's going to be wild. I'm excited for today's conversation. I hope you are too. What's going on, y'all? Welcome to Everyday AI. My name is Jordan Wilson, and this is your daily live stream podcast and free daily news. helping everyday business leaders like you and me,

Starting point is 00:01:40 not just keep up with AI, but how we can make sense of it, get ahead and leverage it to grow our companies and our careers. That's what you're trying to do. Starts right here with the live stream podcast. But if you miss something or if you need more information, it's all going to be in our daily newsletter. We're going to be recapping today's show

Starting point is 00:01:55 as well as all the other AI news you need to know to get ahead. So I'm excited for today's show. Have you not used nanobanana? Don't worry if you haven't. Make sure if you're listening on the podcast. This is one of those more visual ones. So you might want to go to our website and watch the video if you're not going to catch it, you know, catch it on the YouTube stream or anything like that.

Starting point is 00:02:16 But let's just get into it. I'm excited for today's guests. So please help me. Welcome to the show. We have Paige Bailey, the AI Developer Relations Lead for Google DeepMind. Paige, thank you so much for joining the Everyday AI show. Thank you so much for having me. And I can't wait to share more about what we've been building, especially for a nanobanana.

Starting point is 00:02:35 That's the other AI features that we've been incorporating into Google Workspace and our other Google products. Yeah, it's, it's been great. And, you know, as someone that's been covering generative AI every day for like three years, there's been very few moments where I'm like, oh my gosh, there's been like five. And Nanobanana has been one of them. But before we get into it, page, can you just tell me a little bit and our audience a little bit about, you know, what your role is at Google DeepMind and what it is that, you know, you're kind of working on on a day to day. Yeah, so I am the area tech lead for developer experience at Google Deepind.

Starting point is 00:03:09 You might have kind of been familiar with developer relations before. We've kind of expanded this out to include more around creator experience. You know, obviously people are starting to be able to build apps and to build digital artifacts with things that aren't so much code-centric. So being able to kind of help people be successful there to work with the product teams and the engineering teams as well as modeling teams to make sure that we're getting the right data into pre-training, that we have the right e-vowels for the models, and really just trying to make sure that as you're building with Gemini,

Starting point is 00:03:44 as you're building with GEO3, that you're able to do everything that you... Prior to this, I started as kind of a machine learning engineer back in 2009, building models mostly in kind of the Earth's mind background as geophysis and applied math. And then the world has certainly changed quite a bit since then. It's changed a ton, right? So I even remember, you know, I was at, you know, Google Cloud Next. And, you know, you and, you know, Logan took the stage there. And I'm like, I'm like remembering now, like, and just seeing the difference between, you know, what?

Starting point is 00:04:25 That was like April or May and where we are today, what's possible for both developers and non-developers? has completely changed. So just describe for our audience, you know, what are, you know, how capable, you know, are people now to build something that maybe if they didn't have the skills before, how has it changed?

Starting point is 00:04:47 Yeah, it's, I think there's, it's truly remarkable what you can do with Mano Banana in particular. So, so you can think of it as kind of a natural language interface for images that allows you to do most of what you would be doing in Photoshop. So you can remove backgrounds.

Starting point is 00:05:06 You can colorize images that might be historic images of your family. You can make them higher resolution. You can kind of take your Pinterest mood boards and turn them into actual like visualizations using your own home as kind of the background scaffolding. You can create really, really beautiful and hyper-grounded new kind of portrait photos. You can generate your own passport. photos like all sorts of things and it's been really remarkable to see what that people have been doing there and then another product that we're really excited about is called build which allows you to just

Starting point is 00:05:44 describe an app and natural language create it and even deploy it with a unique URL without having to ever write a single line of code yourself so like there's even a feature with jemini that if the if in the process of writing the code for the app um the model encounters any errors it will take the error, feed it back to Gemini, and then kind of self-heal the code and self-heal the app to get the right outcome for what you've described. So obviously, you have a technical background, right? I don't. And what's weird is like everything that you just said there, it all made sense to me,

Starting point is 00:06:22 whereas maybe three to five years ago, some of that would have went over my head. Let me ask you this. Is everyone or can everyone be a developer now, at least? like, you know, go in and get their feet, right? Right? I've used jewels. I've used Oval. I've used, you know, AI studio to build apps, Canvas and Gemini, right?

Starting point is 00:06:42 Can anyone do this or do you still need some level of expertise? Absolutely. Anybody can do this. And I think it's so exciting in the sense that we have hackathons every weekend in the Bay Area or it feels like we do. And it's been really magical to see that the attendees for hackathons have moved away from just being engineers to being planned. to being product managers, to being, you know, folks for maybe a sales background,

Starting point is 00:07:06 a business development background, who are really excited about solving problems, and who are able to describe really articulately like the kinds of presence that need to be solved. And now there's no hurdle. Like there's no big chasm of, you know, software engineering ability that needs to stand between them and getting the work done of building an app that can really meet their customers leaves. So a lot of the time at hackathons, the teams that we see winning and the teams that we see, you know, kind of potentially even building out companies aren't really necessarily the ones with all engineers on staff. It's kind of like the singleton PMs who have a hobby project or the salespeople who have spent, you know, 10,000 hours with customers and who deeply know what they need and who just want to like build a tool that they can help address those concerns.

Starting point is 00:07:57 So I am very jazzed. And I can't wait to see things like build and things like these models be adopted more and more in the science community as well. Because I think we have a ton of hyper-talented people across so many different domains who haven't had the time to spend 10 years learning how to be a software engineer, but who know precisely what needs to be built. And now they finally have AI democratized in a way such that they can just be let loose and create all of the things. things that they've dreamed about. All right. Well, let's let's do this. Let's let's let's let's let loose a little bit. So for our podcast audience, I'm going to have Paige share her screen here. And we're going to walk through a couple of different use cases. One, I think is great for for non-technical people, beginners. If you spend a ton of time in Google sheets like me, I think this first example

Starting point is 00:08:48 will be really, really good. This is something that I love using. But Paige, why don't you just kind of walk and talk us through kind of what we can do here inside of Google Sheets. Let's go ahead and bring it up there. So there we go. So we have your Google Sheet showing. Just walk us through what's going on here. Yeah. So I have a couple of tabs that I've created in this Google Sheets document to help showcase some of the new capabilities of our AI function, which is baking Gemini directly into Google Sheets. This first one is kind of near and dear to my heart. I am a super soccer fan. I used to play it, used to play it back in school. And so we have a lot of teams from the UK, specifically from England, their stadium names and their locations.

Starting point is 00:09:40 And what I'm going to do is in this fourth column, I'm just going to type in equals AI. So that's the function name. I'm going to open parentheses. And then I'm going to describe in natural language something that I would like to have as kind of a supplement to this data. So I'm going to take this stadium names column. I'm going to say something like return the address for the location.

Starting point is 00:10:08 I'm going to add a comma. I'm going to add the kind of link to that cell. So this is all very familiar if you've ever used sheets before, if you've ever used Excel before. I'm going to hit enter. And immediately behind the scenes, Gemini kind of returns the location of the stadium name in the United Kingdom. And then even better, kind of similar for user experience to all of those other functions that you might have used before.

Starting point is 00:10:41 It kind of fills in and populates each one of the cells in the column based on that stadium name and location. So it's pretty cool to be able to see this happen in real time and to be able to understand that, you know, you can start generating data, analyzing data, doing things like sentiment analysis, just be a natural language and the AI function. Like, let me be honest. Like where, like, where was this like 10 years ago, right? Because this one simple thing. So, you know, page walked us through. There's team name, column A, stadium name, column B, location. You know, column C. And then in column D, she just automatically pulled the address live for these 20 top England soccer teams. This would have taken, I don't know, at least for me, two hours because I would have got distracted. I would have went on the team's Twitter. Right. Like that is such a time saver. So even sometimes when you're thinking of like, oh, what would I use, you know, Gemini and sheets for? There you go. Save a ton of time. Exactly. And it's also really useful. Like I'm sure if anybody has ever done data pre-prone. or data cleaning before you've experienced the pain that is, you know, people's names spelled slightly differently or capitalized slightly differently in the same fall on state names, city names, slightly misspelled.

Starting point is 00:12:04 Like, but this function, what you can do is you can kind of go through. And for those fuzzy problems that didn't even have, like, exact functions to help you tackle, this AI feature can kind of go through and make data pre-processing so much easier. I have been using it all the time, honestly. And I think it unlocks kind of these more complicated data analysis workflows for people who previously would have had to learn something like Python or R to do those automations. Yeah. Such like a great, I think, use case for anyone out there, you know, whether you don't want to

Starting point is 00:12:42 have to go through and manually do it or just enrich your spreadsheet. So, all right, what are you showing us here with? Yeah. Yeah. So this is a surprise for, or at least you've probably done it before. So it's, but I haven't shown you this tab. I just grabbed some comments from different social media platforms. So full disclosure, some of the comments might be a little bit spicy because they come from social media. But one of the cool things that you can do with Gem and I as well is you can ask for sentiment analysis.

Starting point is 00:13:13 So we have some gems that are baked in. you can also just ask in natural language to classify based on whatever categories that you look like. But you can kind of give the kind of table views for customer sentiment. You can summarize the customer sentiment of the support tickets. You can kind of analyze at scale how customers feel about a product. Or kind of similar to what we just did before, you could say, like, classify the sentiment for this statement into positive, negative, or neutral. And then just say, like, all right, we'll be two.

Starting point is 00:13:58 And then it will go through and kind of do the classification for you. I've spent, again, all these things, I'm like, like laughing and in like face palming, just hours I've spent. And the money I've spent in years past to pay for tools. that would monitor sentiment analysis on social media and page just showed us, oh, you can just click a button now in Gemini and Sheets. So all right, I want to get into the second use case, but before we do a quick, very quick word, you know, from our sponsors, great timing.

Starting point is 00:14:33 Google, notebook L.M. This podcast is supported by Google. Hey folks, Steven Johnson here, co-founder of Notebook L.M. As an author, I've always been obsessed with how software could help organize ideas and make connections. So we built NotebookLM as an AI first tool for anyone trying to make sense of complex information. Upload your documents and NotebookLM instantly becomes your personal expert, uncovering insights and helping you brainstorm. Try it at notebooklm.com. All right, let's get. I actually literally forgot until just now, if any of you are listening

Starting point is 00:15:11 to this and you're like, wait, I've heard this page sounds so so familiar. I think it's a previous partnership we had with with google i think we had page uh you know talking about jemini 2.5 flash so uh great great timing amazing and i also like i have to say like notebook lm and stephen johnson like he's just such a sweet heart too like they have been uh rolling out features like gang busters it's been so awesome to use audio overviews and all of the analysis features and notebook lm as well oh it's been yeah and and i do have to shout out like josh like i'll touch Woodward, I'll like, you know, DM him something. I'll be like, hey, this isn't working. And then like a day later, he's like fixed it. I'm like, this is crazy.

Starting point is 00:15:54 I don't think he sleeps. Like he is, he has any, he does that for all of the products, but within labs, within deep minds, um, is just kind of like the most helpful person I have worked with in my entire career. So like, if you all don't follow Josh Woodward on all of the social media, I strongly, strongly recommend it. All right. So page, let's let's get into the next one. Let's show people how you can literally build an app in natural language with nano banana, with the viral AI image generator and editor. Let's see it. Walk us through it.

Starting point is 00:16:28 Absolutely. So right now I'm in AI Studio. AI Studio is kind of the first place to go to get access to DeepMinds models as soon as they're released. You can select different models here off to the right, including nanobanana and see kind of information about them behind the scenes. You can also use many of these models in our products like the Gemini app, but this just kind of gives you an under the hood kind of view for a lot of the models and their capabilities. But the feature that we're talking about today is this thing called built. It's this little puzzle piece guy that you might see here off to the left.

Starting point is 00:17:01 When you click on it, you're kind of put into this gallery of a whole bunch of apps that the team is created to kind of inspire you. Or you can just describe an app that you would like to create a natural language. and have it generated for you. So today I am going to, you know, we talked a little bit about Dungeons and Dragons before. So I'm going to just, you know, create an app

Starting point is 00:17:30 that allows all of my friends to generate like hyper-personalized characters using nano-banana. Does that sound good? Let's do it. And literally, like, y'all, there's no editing. You know, if you're listening to the podcast, She's doing this live.

Starting point is 00:17:45 She's typing this out inside of AI Studio in the build feature. So anyone can go do this in natural language. But yeah, maybe why don't you kind of literally just dictate even what you're typing here, page just so our audience can imagine it with you. Yeah, absolutely. So I'm just going to type create an app that uses the webcam to take an image of the user. The app should then use nanobanana to modify the image of the user. into a Dungeons and Dragons character,

Starting point is 00:18:19 a unique Dungeons and Dragons character, make sure that the app also includes character stats, so like strength and dexterity and all that good stuff, and that the app is well designed. So let's do Control Enter and see what we get. That was obviously like a very simple prompt. I'm not the best at prompting. I should have probably used Jimini 2.5 Pro to help me rewrite my prompt behind the seats.

Starting point is 00:18:48 It's better. It's better just to go with your gut something simple. So like as this is building, so it's telling us exactly what it's doing, right? Yeah, it absolutely is. If you've ever played SimCity 2000 and you've seen like the loading screen for like reticulating splines, it looks very, very similar. It's basically defining everything that it would need to do in order to build out this app, with all of the stack associated with building a really nice web application.

Starting point is 00:19:19 If it needs to generate prompts behind the scenes for any of these models, it's kind of defining those prompts as well. We're kind of put into this thing that looks a little bit like a development environment. So you can see the code getting generated off to the right. This really nice directory structure in the center. If you're a developer, you can kind of one button click, save this as a public or a product,

Starting point is 00:19:42 private GitHub repo. And if Gemini encounters any errors along the way as it's building this app, what it will do is it will take that error. It will put it back into the model. And then it will use all of that to kind of self-heal and like fix itself before the app exists. But it looks like we've already got the app. It's done. Yeah, it's done.

Starting point is 00:20:06 So let's test it out. I'm going to begin this quest. It says D&D Character Forge, which sounds. sounds very, which sounds very, very cool. I'm going to take a picture. I'm going to forge my character. They use the webcam to take a picture, so that looks like it was pulled up correctly. It's consulting the elder scrolls and forging arcane artifacts.

Starting point is 00:20:30 This is very cute. So it's got my before image. It's got my after image. That's super cool. I am for anybody, I'm a wood elf ranger, and I'm chaotic good. and it's got like the same, it's got kind of like the same piece sign. It's got the same facial structure. My hair is a little bit purple.

Starting point is 00:20:50 So maybe that's inspiration that I make a lifestyle choice change for like dying. In the future, it seems as well. Exactly. And then it's got strength, dexterity, wisdom, chrysma, all that good stuff, as well as a backstory. So Lyra grew up in a secluded elven forest, learning the ways of nature and the bow from an early age. This is awesome. I love it. This is nuts.

Starting point is 00:21:15 And yes, she literally did this live in real time, just said nano banana. It obviously knew to use Gemini 2.5 Flash. And she used her webcam right there, like just vibe coding in the airport. So now how can you actually use this and share this? Yeah, well, of course, I want you to join my Dungeons and Dragons campaign. So, so like, and all of the folks who are listening today. So I'm going to go ahead and click this little rocket. symbol that we have on the right that allows us to deploy the app automatically to Google Cloud.

Starting point is 00:21:46 I'm going to select the Cloud Project. I'm going to click to deploy the app. And then it will kind of generate a unique URL behind the scenes. So this URL you can share with friends, family, coworkers. And it will also hook up everything else that you need from a cloud project perspective. So storage, logging, cloud run, like it hooks up all of the, all of the infrastructure. so you don't have to worry about any of it. It's doing it in a secure way because it's using Google Cloud. And then if I click this kind of cloud project, it will also give the insight into logs,

Starting point is 00:22:23 so all of the things that were required to build this app, and then also billing. So if this happens to go viral, I can kind of see, and also no judgment, this is my personal GCP account. It shows me all of the services and all of the costs associated for the different products that have been used. So it's really, really nice to have all of this handy. And we were able, like, this is effectively a hackathon project using nanobanana deployed with a unique URL that took us less than like two minutes to build. It's kind of bonkers.

Starting point is 00:23:02 Amazing. Yeah, amazing. Like I, even as, you know, Paige, Paige was talking there, I, I already. used it. I already have my own character. Not as much strength and dexterity as, as Paige, but, you know, I got to start somewhere. I got to start somewhere. All right. So, Paige, you gave us a great, you know, Gemini use case for non-technical people. We dove into and talked about some different nanobanana use cases. You built something literally live in, in like a minute. So I want to ask you this as a third thing. What's been your personal, you know, favorite either use case of nanobanana? You know,

Starting point is 00:23:39 your favorite project you've seen maybe someone else build. What's kind of that one thing that you always go to? You're like, this is such a great use case. Because I feel there's like unlimited use cases. There are so many use cases. One of my favorites has really been because often I don't have time for professional, like, you know, the kind of professional headshots that you might put on LinkedIn for myself.

Starting point is 00:24:02 And I know that my coworkers don't either. You can ask for nanobanana to create professional headshots. I've been using it to colorize old family photos, which has been really exciting for my mom and for all of my family members to see. There's also a really great use case where you can give it an aerial satellite image of a place and then have it turn into an isometric building that you could use for a game. So it looks like kind of a pixelated scene from a video game. You can have it imagine what a street view like it looks like for a given place that you might have on satellite images. But one that everybody on the call can try out today that's part of our kind of Gemini app showcase is actually called Past Forward. It was created by my colleague Amar.

Starting point is 00:24:57 And what you can do is you can generate yourself through the decades, which is quite cool. So if I was to kind of upload a photo, so I'm just going to select an old photo that I have kind of hanging out on my, hanging out on my laptop. I'm going to click generate. And then this app behind the scenes, and I'm going to zoom out a little bit so folks can see, it creates different images from the 1970s, from the 2000s. And kind of allows you to kind of see yourself in all of those different age. I love the aviator glasses from the 70s, like holy moly. But it and also the big pair from the 1980s.

Starting point is 00:25:46 Like I would absolutely rock that. Like this is this is really, really cool. And is something that you could do today with your friends, your family and kind of have displayed along the way as well. the possibilities are literally mind-boggling, right? Absolutely. Paige, I could have you show us hundreds of demos every day, and I still don't think we could honestly scratch the surface of what's possible in nanobanana.

Starting point is 00:26:15 But, you know, as we wrap up today's show page, what would you say is your one, you know, most, you know, practical or tactical, you know, piece of advice that you have for people to start going in and maybe either using, nanobanano on the front end in AI studio or building with it. What's that one piece? Yeah. So I would strongly, strongly recommend. I know there's a lot coming out.

Starting point is 00:26:41 Like deep mind, it feels like we're releasing new models and new features every four or five days. So it's been a hard thing to keep up even for all of the folks working on the team. But really the best way to kind of understand what these models are capable of is to test them out yourself. So roll up your sleeves like Google's made a lot of these models available to use for free via the Gemini app via NoBook Gillem via AI Studio. So if you have something that you've been curious about, if you have like a workflow that's really frustrating that you would love to automate, just go in, ask Gemini to help you to help you write a prompt if you don't

Starting point is 00:27:24 know how to prompt yourself or if you're just getting started and just test it out. And I guarantee you if you try it, if you keep iterating, if it doesn't work the first time, I think that you will be surprised to see how far these models have come over the last six months. And given that this is a new space, given that this is a complete reimagining of everything, machine learning and AI over this past couple of years, like you're getting started at pretty much the same place that everybody else is getting started. So really, there's still time, there's still room to get in on the ground, or don't be afraid, like, just test things out.

Starting point is 00:28:02 I don't know what's more impressive, you know, what we've seen in Nanobanana or what Paige Bailey just delivered in 27 minutes in terms of value. My gosh. So if you miss anything, if you're listening, make sure you go check out our newsletter. We're going to be recapping it all. But Paige, thank you so much for taking time out of your day to join the Everyday AI show. We really appreciate it. Thank you.

Starting point is 00:28:23 Thank you for everything that you're doing to kind of bring AI to everyone and to make it approachable and accessible. We appreciate you and I can't wait to listen to your next podcast. All right. Well, hey, every day. So you can't, it's hard to miss it. It's hard to miss it. All right. So thank you everyone for tuning in. Like I said, if you haven't already, please go to your Everyday AI.com. Sign out for that free dline newsletter. We'll see you back later for more everyday AI. Thanks, y'all. Awesome. Meet Firefly AI assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps,

Starting point is 00:29:05 including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us.

Starting point is 00:29:36 us going. For a little more AI magic, visit Your EverydayAI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.

Everyday AI Podcast – An AI and ChatGPT Podcast - EP 619: Nano Banana Uncovered: A practical guide from inside Google with Paige Bailey

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.