Unlock the secrets of the Google Nano Banana Upgrade! 🚀 In this video, we're diving deep into the **fastest and most effective ways** to enhance your trading experience using cutting-edge AI tools. You’ll discover tips and strategies to maximize your gains and easily navigate the digital landscape! 📈💰
Whether you're a seasoned trader or just starting out, this upgrade will change your approach to online income and crypto trading. Get ready to elevate your skills and achieve your financial goals!
**Registration Link:** [https://Tava.short.gy/aibanana](https://Tava.short.gy/aibanana)
Don’t forget to like, comment, and subscribe for more exciting insights and updates! 🔔
#NanoBanana #CryptoSecrets #TradingTips #AIUpgrade #FinancialFreedom #OnlineIncome #InvestSmart #CryptoTrends
Whether you're a seasoned trader or just starting out, this upgrade will change your approach to online income and crypto trading. Get ready to elevate your skills and achieve your financial goals!
**Registration Link:** [https://Tava.short.gy/aibanana](https://Tava.short.gy/aibanana)
Don’t forget to like, comment, and subscribe for more exciting insights and updates! 🔔
#NanoBanana #CryptoSecrets #TradingTips #AIUpgrade #FinancialFreedom #OnlineIncome #InvestSmart #CryptoTrends
Category
📚
LearningTranscript
00:00So let's create an image, give it a prompt here,
00:02and click submit, and it's using Nano Banana 2.
00:05Let's see how quickly this works, and it's done.
00:08So Nano Banana 2 is live,
00:10and it's supposed to give you
00:11Nano Banana Pro-like intelligence and quality,
00:14but at flash speed.
00:16So in this video, I'm gonna break down
00:18what's new about Nano Banana 2,
00:19and then we'll test all the biggest claims.
00:22Speed, text accuracy, translation, subject consistency,
00:25instruction following, and 4K output.
00:28As always, I'm not here to waste your time.
00:29Let's just get right into it.
00:30We just generated this photo realistic
00:32matte black reusable water bottle,
00:35and it did it in just a few seconds.
00:37Let's toss a logo in here and say,
00:39add this logo to the water bottle.
00:42Should be pretty dang quick.
00:44And while I did speed that up,
00:46it literally only took about 10 seconds or so.
00:48Let's do one more iteration on this same image.
00:51Make the same image, but with dramatic nighttime lighting
00:53from a desk lamp, keeping the composition and realism.
00:56No text.
00:57And yes, I did speed this up again,
00:58but it only took about 10 seconds,
01:00and we've got this new iteration.
01:03So what is Nano Banana 2?
01:05This is Google's latest state-of-the-art image model,
01:08and it's also known as Gemini 3.1 flash image,
01:12but Nano Banana 2 is a lot easier to say.
01:15Basically, the pitch of this new model from Google
01:17is that it is pro-level quality and intelligence
01:20at flash-level speeds,
01:22and so far, it seems to be pretty dang fast.
01:25But that's not the only things that are new about it.
01:27So let's break down what's actually new here.
01:29These are the claims from Google
01:31that we will actually test in this video.
01:33It's got that same advanced world knowledge
01:35and web grounding that Nano Banana Pro has,
01:38but at the flash speeds.
01:40It's got accurate, legible text and translation
01:43at the flash-level speeds.
01:45The same level of subject consistency.
01:47You can have up to five characters
01:49and up to 14 objects all being added into a single image.
01:53It supposedly has a much stronger instruction following
01:57than even Nano Banana Pro did.
01:59You can get outputs in production-ready formats
02:02all the way up to 4K,
02:04and they're supposedly a visual fidelity upgrade
02:07while staying really fast at these flash-level speeds.
02:11Now, you can access this inside of your Gemini account.
02:13You go to gemini.google.com,
02:15click on Create Image,
02:16or select Create Image from the little dropdown here,
02:20and you can instantly tell
02:21you're using the new version
02:23because you have all of these styles available to you now.
02:27If you select any one of these styles,
02:28so for instance, Gothic Clay,
02:31it will use this as a reference image down here
02:34to generate it in that style.
02:35So let's do a wolf howling at the moon,
02:37and I should get it in this Gothic Clay style,
02:40and it should do it pretty dang fast,
02:42because once again, it's using Nano Banana 2,
02:44which is using the Gemini 3.1 flash model.
02:47And a few seconds later,
02:48I've got a wolf howling at the moon that looks straight out of the Nightmare Before Christmas.
02:52Now, I was given early access to Nano Banana 2,
02:54and they gave me that early access inside of the Gemini app,
02:58but it is available on other platforms.
03:00I just didn't get early access to those to record as part of the video.
03:03It will also be available inside of AI Studio.
03:06So by the time you're seeing this video, you can use it over there.
03:10Go to AI studio.google.com.
03:12Up in the top right, where you see the model,
03:15go ahead and click on that model, select images.
03:17And by the time this video is live, you should see an option for Nano Banana 2.
03:21Although again, I don't have access in this platform yet.
03:25It's also available in Google Cloud slash Vertex and inside of Google Flow.
03:30It will now actually be the default image generator inside of Flow.
03:33It actually uses zero credits if you use it inside of Flow.
03:37But again, my early access was sort of generalized to the Gemini platform for me to test.
03:42If you're in Gemini, you can use Nano Banana 2 for free, as far as I'm aware,
03:48in about 141 different countries.
03:50The Nano Banana Pro model that we used to have access to is now going to be on the Pro
03:56and Ultra
03:57plans.
03:57So on one of the paid plans.
03:59And if you are on one of those paid plans, like Pro or Ultra, and you want to actually upgrade
04:04this
04:04image, well, you can actually turn this into a Nano Banana Pro image by clicking on these three
04:09dots and clicking on redo with Pro.
04:12So if you don't like Nano Banana 2 and you prefer to use the older Pro model, you can on
04:18their paid plans by generating with Nano Banana 2 first and then re-rolling it as a Pro image.
04:23All right, now let's get into some real tests with this.
04:26So let's do a little side-by-side comparison here.
04:29On the left, I'm using the fast model.
04:31On the right, I am currently using the Pro model.
04:33So I'm going to put the same prompt into both of them.
04:36We'll start the timer and see how it really compares.
04:39So here we go.
04:40Here's our stopwatch.
04:41Again, Nano Banana 2 is on the left and the Pro is on the right.
04:45And okay, so the first one is done and now the second one is done.
04:49And when we compare them side-by-side, I mean, they're pretty close in quality.
04:53I would say this one on the right is still a little bit more high quality.
04:57It looks a little more realistic to me where the one on the left looks a little more AI
05:01generated, but they're pretty dang close.
05:03And the one on the left was definitely quite a bit faster.
05:06Now let's have them both edit the same image.
05:08So again, fast on the left, Pro on the right, submit, submit, and starting my stopwatch here.
05:14Now I sort of jump cut into the final result, but I was looking at the stopwatch.
05:18The image on the left, the fast image took 13 seconds.
05:21The image on the right took 29 seconds.
05:24So more than double the time to generate the edit on the right.
05:28I'm going to try one more simple prompt here.
05:30I'm going to ask it to change the bandana to red.
05:32The first one's done at 15 seconds this time.
05:35And the second one's done at 34 seconds.
05:38So again, more than double the amount of time to do it on the right as the left.
05:43Now, again, I would say Pro is still slightly higher quality, but it's pretty dang close.
05:48All right.
05:48Now let's test text rendering.
05:50So again, I'm going to test them both side-by-side.
05:52Make sure my right side here is on Pro, left is on fast.
05:56I'm going to stop doing my timer because what we know is that Pro takes about double the time
06:01as fast and fast usually falls within the 13 to 15 second range to generate,
06:07where Pro tends to fall in the 25 to 35 seconds to generate.
06:11So for text rendering, I'm going to give both of them a pretty in-depth prompt here.
06:15Design a photorealistic laptop on desk scene with a browser window open to a pricing page
06:20for a fictional product called Banana Studio.
06:22So requirements, headline, Banana Studio pricing, subhead, choose a plan that fits your workflow,
06:27footnote, line with an asterisk, a three column comparison table with headers, starter, pro,
06:32team, three table rows labeled exactly projects, image exports, support.
06:36Fill the table with these exact values, a bunch of values for the price,
06:40fine print at the bottom in small text.
06:42I'm giving it some extra rules.
06:43All text must be perfectly spelled, readable, and aligned.
06:47No extra words, no gibberish.
06:48Character's clean, modern UI design.
06:51Both are getting the exact same prompt.
06:53Not only is this going to test actual text rendering in the images,
06:56it's going to test the ability to follow complex instructions
07:00and we can see how many of the instructions it actually got right.
07:03All right, so here's our side by side.
07:05We can see on the left here, here's what it generated.
07:07Let me open up the prompt here again on the right so we can actually see it.
07:10So Banana Studio, headline, Banana Studio pricing, subhead, choose a plan that fits your workflow,
07:17annual billing saves 20%.
07:19Here's our columns.
07:20So we got our three column comparison, starter, pro, team.
07:24I literally see no mistakes.
07:26So what does that say down in the bottom?
07:28Prices in USD, taxes may apply, terms and conditions apply.
07:31It nailed it.
07:32It literally got every single instruction that I put in there.
07:36All right, let's compare it to this one here.
07:38I mean, they're on par.
07:39I wouldn't really say the one on the left or the right is better.
07:42If anything, I actually like the look of the laptop on the left a little bit better,
07:46but both nailed all of the text and all of the instructions that I gave it.
07:50Except the one on the left, you could use for free and it only takes about 13 to 15 seconds.
07:56And the one on the right, you need a paid plaid to use and it takes like 35 seconds.
08:00All right, next up, let's test translation and localization within an image.
08:05First, I'll have it create an image in English and then get it to translate that image.
08:09Again, we'll do fast on the left, pro on the right.
08:11I'm going to give it a prompt to make a modern event poster with a bunch of details in English
08:17here.
08:18Going to do the same on the right over here and submit them both.
08:21If you want to pause right here, you can actually see the entire prompt.
08:25And here's what fast gave us while pro is still working.
08:28Creator workshop, make better videos with AI.
08:31March 12th, 630 PM downtown community hall, free entry.
08:34And here's the one that pro made.
08:36Everything looks good, but I'd actually argue that the one on the left, it looks better.
08:40There's more color, more details.
08:42I mean, it's a more impressive looking image to me.
08:45All right, now let's get them both to translate these posters to Spanish.
08:48So I'll give it the prompt, edit the poster, translate all text into Spanish while keeping
08:52the exact same layout, typography, style, and spacing.
08:55Give the same prompt to pro over here and we'll submit both of these.
08:59And a few seconds later, we've got our Spanish translation.
09:03And both seem to have done a pretty good job.
09:05The only thing I'm uncertain about is downtown the same in Spanish as it is in English.
09:11Because if it isn't, then this one did it right.
09:14I think this is the more accurate translation here.
09:18So fast actually did the translation better than the pro model.
09:22All right, next up, I'm going to test subject consistency, but this time I think I'm just
09:26going to get rid of the pro model and we'll just test in the fast.
09:29We know that they're relatively similar.
09:31We're getting pretty much the same kind of results, sometimes even better out of the
09:36fast that we are on the pro.
09:37So let's just play around with the fast model from here on out.
09:40So I'm going to give it this long prompt here where it's going to generate five characters
09:44for us.
09:45An adult woman with curly black hair and a green sweater.
09:48An adult man with short blonde hair and a gray hoodie.
09:50A teen girl with long brown hair and a yellow beanie.
09:53A teen boy with short red hair and a denim jacket.
09:55And an elderly man with white hair and a navy cardigan.
09:59Now I'm generating this first prompt, not really to see how well it does these people.
10:03I'm generating this first prompt because I want to give follow-up prompts and see how
10:07well it maintains the same characters that it made in the first prompt.
10:11So let's go ahead and generate this as our starter here.
10:14And we've got our initial image.
10:16An adult woman with curly black hair and a green sweater.
10:19Check.
10:19Adult man with white short blonde hair and a gray hoodie.
10:23Check.
10:23Teen girl with long brown hair and a yellow hoodie.
10:26Check.
10:26Teen boy with short red hair and a denim jacket.
10:30Check.
10:30And an elderly man with white hair and a navy cardigan.
10:34Check.
10:34Okay.
10:34So we've got all of our characters here that we asked for.
10:37Now let's create an image and see if it'll create an image with the exact same characters,
10:41but in a different scenario.
10:42And here are our action changes.
10:44The teen boy picks up the skateboard.
10:46The adult woman is standing and taking a sip from the red mug.
10:49The orange cat is now on the wooden coffee table.
10:52I gave it some extra rules here.
10:54All characters must look like the same people as frame one.
10:57All 14 objects must still be present and recognizable.
11:00I guess I probably should have mentioned that.
11:01I also had to do these 14 objects inside of the image
11:05so that we could make sure those objects stay consistent as well.
11:08So let's see how well our scene change looks.
11:11And here's the new image.
11:12So the woman is standing, taking a sip.
11:15The cat is on the table.
11:16The boy picked up the skateboard.
11:17If we take a look at these two images side by side here,
11:20it kept the characters really consistent.
11:22Everything looks the same.
11:24And then our 14 objects, our red mug, a blue notebook, a silver laptop,
11:28a black smartphone, a green house plant, our orange cat, wooden coffee table,
11:32our floor lamp, a framed mountain photo on the wall, a TV remote.
11:36We have our pair of sunglasses, which seems to have been removed on the right image.
11:42So that's one thing that got removed, even though the instructions said not to remove it.
11:47When the cat jumped on the table, it must've sat on the sunglasses.
11:50So I guess we could just maybe assume the sunglasses are under the cat here.
11:53We have our skateboard and we have a small cactus in a white pot.
11:57And that actually is on there twice, but everything is there.
12:01The only thing that was removed was those sunglasses.
12:03All right, let's generate another frame.
12:05This time we'll do a camera change.
12:07Switch to a wider angle from the opposite corner of the room.
12:11Keep the same lighting time of day, same character identity, same wardrobe.
12:15All 14 objects are still present and recognizable.
12:18Let's go ahead and see if it can do that for us.
12:20Get a different angle while maintaining the same characters.
12:24Well, that's not looking good.
12:25It didn't change the angle for us.
12:27If we take a look at these two images here, it changed a few things.
12:31The cat moved.
12:32The boy is now turned and looking out the window.
12:35The elderly man is now looking towards the woman instead of down at his paper.
12:39And the sunglasses are still gone, but it did not change the camera angle for us.
12:43Maybe I could be more explicit about that and just give it the prompt.
12:47Show us the same image, but from the angle of the opposite corner.
12:50Okay, so we got a new angle here, but it didn't keep consistency of the room.
12:55That's for sure.
12:56So if we take a look at these two side by side, it did keep the characters consistent.
13:00So it's got that going for it.
13:02It definitely will keep the characters consistent for you.
13:04The redheaded boy is now gone completely.
13:08The elderly man is in a different spot.
13:10The blonde man is in a different spot.
13:12The girl is kind of in the same spot, but also kind of moved.
13:16And the cat looks like, you know, maybe it would be from the other angle.
13:19But this is a really weird layout for an apartment.
13:21If there's a window that looks identical on the opposite corner.
13:26So yeah, it didn't quite get the rearranging.
13:29We can also see the lamp is in the exact same spot,
13:31even though it's supposed to be from a different angle.
13:33Rearranging the angles, maybe still some issues here.
13:36Now let's do a little more testing on instruction following.
13:39Now we did a test with the laptop and it got all of the text that we asked it for.
13:44So we already know it's pretty good.
13:45And that last one, I asked it to put 14 images and five characters in there,
13:49and it followed all of those instructions.
13:51Got a little wonky when we started rearranging things,
13:54but it did follow all the instructions on the initial prompt.
13:57Let's test one more and see how well it follows these instructions.
14:00So create a photo realistic studio product photo of a single pair of matte black over
14:06ear headphones on a pure white background.
14:09Headphones centered, perfectly symmetrical, soft shadow directly beneath, no hard edges,
14:14no logos, no branding, no text anywhere, no extra objects, no cables, no stands,
14:19clean realistic materials, subtle texture on ear pads, 85 millimeter lens look with a 5.6 F stop,
14:26crisp detail, lighting stop box from upper left, faint fill from right, aspect ratio 1.1.
14:33Let's just make it a square.
14:34Okay.
14:34Headphones centered, perfectly symmetrical.
14:36I mean, yeah, that I'd say that's pretty symmetrical, soft shadow directly beneath.
14:40Yeah.
14:41No logos, no branding, no text anywhere.
14:43Yep.
14:43No extra objects or cables or stands, subtle texture on ear pad.
14:47We take a closer look.
14:48We've got that subtle texture on the ear pads that you'd expect and the lighting,
14:52the soft box from the upper left.
14:54So if we imagine the light source is coming from up here, shining onto our headphones,
15:00I would say that is kind of the direction the shadow would cast.
15:03So yeah, I think it got all of that pretty good.
15:06All right.
15:07Now let's see if we can get it to rotate the headphones, rotate the headphones,
15:0915 degrees to the right, keeping everything else identical.
15:12No new objects, no text, no logos.
15:14And yeah, I think it did a good job.
15:16It just made a slight rotation to the right.
15:18All right.
15:19For my next test, I want to test some different aspect ratios and really sort of
15:22check out the level of quality.
15:24So I'm going to give it this prompt here.
15:26Let's see what we get out of it.
15:27It's 16, nine, but I actually want to download the full size image and sort of zoom in on
15:32the quality a bit.
15:33So here's our image at full screen.
15:35Let's go ahead and zoom.
15:37This is fully zoomed in right here.
15:39And I mean, yeah, the details are pretty good.
15:42It doesn't look too pixely or grainy.
15:44It looks like a pretty good high quality image to me.
15:47Now I published these videos in 1080, so you're not even going to be able to see it at
15:51its full quality.
15:52But from my eyes, it looks really, really good.
15:54All right.
15:55Now let's get it to generate a different aspect ratio.
15:58Recreate the same scene as a vertical composition for a short form video cover.
16:02Keep the same subject and lighting, but reframe it to 916 with the neon sign centered.
16:07And yeah, that looks good to me.
16:08Let's open it here.
16:09It's the same image, just recompositioned for 916.
16:13Let's give it pretty much the same prompt, but make it four or five this time.
16:17Yeah, looks good to me.
16:18Did it exactly like I would have expected it to do it.
16:20Now let's see if I can get it to generate in 4k.
16:23Because when I downloaded the very first image, if I look at the info for the image,
16:27it actually generated it at 2752 by 1536.
16:33So not quite 4k.
16:35So let's see if I can force it into 4k.
16:37So in order to do that, I'll explicitly ask it to generate in 4k.
16:41Generate this at the highest available resolution, target 4k, prioritize sharp edges,
16:46clean gradients, realistic materials, and no artifacts.
16:49And I'm going to remind it 16, nine aspect ratio.
16:52So it is using nano banana two for this.
16:55Let's see if it will actually give me 4k this time.
16:57All right.
16:58So out of the gate, it looks good.
16:59Let's see if I download at full size, if it actually downloads at 4k.
17:04And once again, it downloaded at 2752 by 1536.
17:08So I don't know exactly how to get 4k out of this, even though it seems like supposedly it can
17:14do 4k.
17:15Now just out of curiosity, if I come over here and redo with pro, let's see what this looks like.
17:21This time it's loading nano banana pro.
17:23Okay.
17:23So this is the one that it generated with nano banana pro.
17:26If I flip back and forth here, you can see they look slightly different.
17:29Like pro is slightly more zoomed in, but they're fairly similar.
17:33Let's go ahead and flip over to the pro version and I will download full size on this one.
17:38And once again, it generated in 2752 by 1536.
17:43So I don't know how to get it to generate in 4k.
17:46It should be 3840 by 2160, I believe to be true 4k.
17:52And I'm not quite getting that.
17:53So again, I don't quite know how to get this to generate in 4k, even with nano banana pro,
17:58but they're still looking pretty high quality, even at the resolution they're at.
18:03There's one final test I want to try out here.
18:05So let me go ahead and create a new chat.
18:07Let's test its ability to find actual world knowledge and
18:11ground with information that it finds on the web.
18:14I tested this when nano banana pro came out.
18:16So let's test it with nano banana to create an infographic of petco park in San Diego and
18:21point out important landmarks around the ballpark.
18:24So let's see if it will actually do the research and understand what petco park
18:29supposed to look like and what to point at and various things like that.
18:34In the past, I gave it an image of petco park and had it point out things.
18:37So let's see how this one did.
18:39So yeah, I mean, it's not a proper layout.
18:42The San Diego convention centers on like the water side of petco park, not the opposite side.
18:48And the park at the park.
18:49Gallagher square is actually inside the park.
18:52So I mean, the layout isn't quite right.
18:55But it did find actual areas like it did name things that are near the park.
19:00So I'll grab this image off of Google images here.
19:03Drop this in and say annotate this image with the landmarks around the stadium.
19:10Let's see if I can figure this out.
19:11Okay, actually not bad.
19:12So the Western Metals supply building is not this building here.
19:16So that's not pointing technically to the right spots.
19:19The Western Metals building is over here.
19:21This is the Omni Hotel San Diego.
19:23So it got that right.
19:24This area kind of over here is all the gas lamp quarter, but it puts it way over here.
19:29This is the Manchester Grand Hyatt.
19:31I do think that's the Hilton San Diego Bayfront.
19:33So, I mean, it did a decent job.
19:35Not all of it's perfect.
19:36It did get Coronado Island out here.
19:39Having some time to have played around with this,
19:40I did get slightly early access to this and had it for a few days before it went live.
19:45My final thoughts are it's really good.
19:47Like I think for the most part, you can use this one over Nano Banana Pro for like almost
19:51all use cases.
19:53The image quality is pretty much as good as what you're going to get out of Nano Banana Pro.
19:56Instruction following is pretty spot on.
19:58Character consistency, pretty spot on.
20:01Translation integrity, pretty spot on.
20:03The only areas that I feel like Nano Banana Pro is still a little bit better at is I would
20:07say if you're going for ultra realism,
20:08Nano Banana Pro to me still looks slightly more realistic in what it generates.
20:14And also, if you're going for the like search grounding, like actually finding information
20:18and making sure all of that information that it went and found is in that image,
20:23Nano Banana Pro still seems a little bit better at that to me.
20:26But for like 95% of use cases, it seems like this Nano Banana 2 is going to be the
20:33daily driver
20:33for most people.
20:35Again, Nano Banana Pro is only going to be in the paid plans.
20:38But from my experimentation, there's very few use cases where you're actually going to
20:43need to switch to Nano Banana Pro.
20:45Like maybe if you wanted to make an infographic where it goes and researches what goes into
20:49that infographic first and then puts it in there, it might do a little bit better job at that.
20:53But even saying that the fast does a pretty good job.
20:57Like if I was to go back and forth and prompt this into submission to where it finally did exactly
21:02what I was looking for, I could probably get this to do what I wanted.
21:05It just might take me a few extra prompts.
21:07So yeah, that's what I got for you.
21:08Nano Banana 2 is out.
21:10It is pretty much as good as Nano Banana Pro.
21:13I wouldn't say it's a huge leap in capabilities beyond Nano Banana Pro, but it's way faster.
21:19I mean, it's twice as fast.
21:20The generations that are coming out of it are pretty much just as good as what you're getting
21:24out of Pro.
21:24It's got some cool new style templates that you can start from.
21:28So if there's a specific design you're going for, you can just use one of them that already exists.
21:33And it's pretty much available for free everywhere in like all of the Google platforms.
21:39Pretty cool upgrade in my opinion.
21:40Anyway, that's what I got for you today.
21:42If you like staying looped in on all the latest AI news and learning about the coolest AI tools
21:47that are available, I got you covered.
21:49Maybe consider liking this video and subscribing to this channel.
21:51I will keep you completely tapped in.
21:53And every single Friday, I do an entire breakdown of all the AI news and cool tools that came out
21:58throughout the week.
21:59So if you just want to watch one video a week and get totally looped in on everything that's happening,
22:04definitely subscribe to this channel.
22:05Anyway, thanks again for nerding out with me.
22:07I really, really appreciate you hanging out and trying new stuff with me.
22:10And thanks again to Google for giving me early access to this.
22:14This isn't sponsored by Google.
22:15I just had a little bit of extra time to play with it, so thought I'd make a video about
22:19it.
22:19So thanks again to them and thanks again to you for tuning in.
22:22And yeah, that's what I got.
22:23All right, bye.
Comments