Экстракт Знаний

STOP Prompting Blindly: The "Nano Banana" Method for Perfect AI Images

8:14

STOP Prompting Blindly: The "Nano Banana" Method for Perfect AI Images

Teacher's Tech 15.12.2025 4 712 просмотров 200 лайков обн. 18.02.2026

Смотреть на YouTube

Поделиться Telegram VK Бот

Транскрипт Скачать .md

Анализ с AI

Описание видео

Are you tired of AI image generators ruining your pictures with garbled, unreadable text? Are you frustrated by characters that change completely every time you hit "generate"? Most people are prompting blindly, crossing their fingers and hoping for a good result. It’s time to stop guessing. In this video, I break down the "Nano Banana" method—a structured approach using Google Gemini’s latest image generation model to finally get perfect text rendering, consistent characters, and photorealistic product mockups. I’ll share my 4-part blueprint (SUBJECT + ACTION + CONTEXT + ART STYLE) that turns vague ideas into precise imagery, and walk you through live examples ranging from cyberpunk cities to instant branding mockups. ⏰ TIMESTAMPS: 0:00 Intro 0:35 Tech Note 0:54 The 4-Part Perfect Prompt Formula 1:15 Example 1: Achieving Character Consistency 2:28 Example 2: Text & Logos 3:33 Example 3: Logic & Infographics 4:13 Example 4: Style Blending 4:53 Example 5: In-Painting (Fixing Mistakes) 6:11 BONUS: Creating Instant Product Mockups 7:31 Conclusion & Final Tips 🧠 PROMPTS USED IN THIS VIDEO: Here are the exact prompts used in the examples. Copy and paste these into Gemini to try them yourself! The Formula: SUBJECT + ACTION + CONTEXT + ART STYLE ► Example 1: A close-up portrait of a woman named Sarah with curly red hair, wearing a green innovative tech hoodie, in a modern office, cinematic lighting. ► Generate an image of Sarah from the previous image, now hiking a mountain trail, wearing the same green hoodie, wide angle shot. ► Example 2: A die-cut vinyl sticker design of a cute robot barista holding a coffee cup. The coffee cup says 'Jamie's Brew' in bold, readable, bubble letters. White border, flat vector style ► Example 3: A clean, modern infographic illustrating the water cycle. Include arrows showing evaporation, condensation, and precipitation. Use a flat design style with a blue and white color palette. ► Example 4: Reimagine this image as a vintage 1920s charcoal sketch. High contrast, rough paper texture, keep the facial structure. ► Example 5: In-Painting (Adding an Object) (First, generate the base image): "A photograph of a clean, empty wooden desk surface with no objects on it, situated in a sunlit home office during the day, with soft natural light coming from a window to the left. Realistic photo style." (Then, use this prompt to add the object): "Add a silver open laptop here displaying a spreadsheet." ► BONUS: Product Mockup (Mug & Logo) (You need two source images first. Image 1: A blank mug. Image 2: A vector logo with a transparent background.) (Upload both, then use this prompt): "Combine these images. Place the vector logo from the second image onto the front of the white coffee mug from the first image. The logo should curve realistically around the mug's surface and follow the studio lighting, showing highlights and shadows consistent with the ceramic material."

Оглавление (10 сегментов)

0:00 Intro 132 сл.
0:35 Tech Note 51 сл.
0:54 The 4-Part Perfect Prompt Formula 78 сл.
1:15 Example 1: Achieving Character Consistency 207 сл.
2:28 Example 2: Text & Logos 194 сл.
3:33 Example 3: Logic & Infographics 109 сл.
4:13 Example 4: Style Blending 124 сл.
4:53 Example 5: In-Painting (Fixing Mistakes) 256 сл.
6:11 BONUS: Creating Instant Product Mockups 252 сл.
7:31 Conclusion & Final Tips 147 сл.

Intro

Have you been hearing the term nano banana lately in the AI world? It sounds like a smoothie flavor, but it's actually an internal code name for one of the most powerful image generation updates we've seen from Google Gemini yet. If you've noticed that your AI images suddenly have better lighting, readable text, and actually follow your instructions, you're likely using one of these new models. Today, I'm going to show you how to get the most out of this update. We're going to look at the best way to structure your prompts. I'm going to walk you through five specific examples from creating consistent characters to generating accurate text and logos. First, a quick tech note. When we talk about Nano Banana, we're usually referring to the Gemini 3. 0

Tech Note

Pro or the updated 2. 5 flash image capabilities. The big changes here is that the pictures just don't look prettier. It's that the AI now has better instruction following. It listens to the complex logic much better than before. But even with a smart model, you need a blueprint. I

The 4-Part Perfect Prompt Formula

always stick to this four-part formula. One is the subject. Who or what is it? Two, action. What are they doing? Three, context. Where are they? What's the lighting? Four, art style. Is it a photo, a 3D render, or a sketch? Keep that formula in mind. And let's jump onto the computer and see this in action. Here I am in Gemini. And we can quickly turn to Nano Banana. You can see I can just click

Example 1: Achieving Character Consistency

create image here. Or when I click tools, I'm going to click create images. Then we get the ban banana up. So the number one question I get is how do we keep the same character across different images. With this new model, this is much easier. Let's try to generate a specific character. First, I'm going to give it this prompt. A close-up portrait of a woman named Sarah with curly red hair wearing a green innovative tech hoodie in a modern office cinematic lighting. I'm giving her a name, a specific look, and a distinct green hoodie. Let's hit enter. Okay, we have Sarah. Now, traditionally, if I asked for a new picture, the AI would give me a totally different person. But watch this. I'm going to give it this prompt. Generate an image of Sarah from the previous image, now hiking a mountain trail, wearing the same green hoodie, wide-angle shot. I'm explicitly referencing Sarah from the previous image. Let's hit enter. Look at that. The facial features are retained. The hair texture is the same and the hoodie is consistent. If you're trying to make a storyboard or a children's book, the consistency is a gamecher. Next, let's talk about text. AI

Example 2: Text & Logos

used to be terrible at spelling. It would give us gibberish. The nanobanana models have solved a lot of this. Let's go ahead and give this a try. I'm going to go to my nanobanana. And let's say I need a sticker design for a coffee shop. I want to have specific text on it. And I'm going to give this prompt. A die-cut final sticker design of a cute robot batista holding a coffee cup. The coffee cup says Jaime's Brew in bold. Readable bubble letters. White border flat vector style. I'm going to hit enter. Notice I asked for bold readable letters. And there it is spelled correctly. Jaime's Brew. The ability to render text means you can use this for YouTube thumbnails, logos, a presentation slides without having to open Photoshop or fix the spelling errors. The one tip I would give is be specific on your text. Tell Gemini Nano Banana exactly what to write. I find when it does make mistakes is when you're telling it to create the text and you haven't been specific on it. The third example is for teachers and business professional. It's about world

Example 3: Logic & Infographics

knowledge. The AI understands how things work, so we can ask for diagrams. Let's ask for something educational. We're going to use this prompt. A clean, modern infographic illustrating the water cycle. Include arrows showing evaporation, condensation, and precipitation. Use a flat design with a blue and white color palette. Now, we're going to hit enter. It usually gets the flow correct. Arrows pointing up for evaporation, down for rain, and it's just not painting a picture. It's visualizing a concept. This is perfect for slide decks where you need a custom graphic that doesn't exist on stock photo sites. Now, for something creative. Sometimes you have a photo,

Example 4: Style Blending

but you want it to look like a drawing or a clay or a cartoon. We can use the upload feature for this. I'm going to upload a photo of myself. And now I'm going to prompt it to change the style, but keep the structure. This is my prompt. Reimagine this image as a vintage 1920s charcoal sketch. High contrast rough paper texture. Keep the facial structure. I'm going to go ahead and submit. It kept the pose and the composition but completely change the medium. You can do this with climatian style pixel art or even oil painting. It's a great way to make your profile picture stand out. Finally, let's look at localized editing. Sometimes you get a nearly perfect image,

Example 5: In-Painting (Fixing Mistakes)

but it's missing one key element. Here I have a generated image of a desk, but it's totally empty. I want a laptop right here in the middle. Now, the great thing about this new model is that I don't need a complicated lasso tool. The AI understands the context of the scene. I just need to tell it what to add and it will figure out the best place to do it. So, in the prompt box, all I'm going to say is add a silver open laptop here displaying a spreadsheet. I'm going to hit enter. By saying add here, the AI looks for the most logical open space on the desk surface. Look at that. It analyzed the lighting and perspective and it inserted the object perfectly. This saves so much time compared to regenerating the whole image hoping for the right result. If you want to control exactly where you want something added, try this. So, I just uploaded the same image here and I'm going to click on it. And notice we have a few tools in here. I'm just going to go ahead and make a circle right here. And I'm going to click done. Now, what I'm going to do is add this prompt. Place a parrot where I circled and remove the circle. I'm going to go and hit enter. And look at this. We now have the parrot on the ledge and the circle removed. I have one more bonus tip for you. And this one is

BONUS: Creating Instant Product Mockups

huge for anyone doing branding or print on demand. We've talked about generating images from scratch, but what if you want to place your own design onto a real product? Let me show you how to create an instant photorealistic mockup. I have two images here. First, the perfectly blank white ceramic coffee mug, and second, a retro coffee logo design. I want the AI to wrap it around and curve on the mug and apply realistic lighting to it. I have the two images uploaded into chat. And here's the prompt I'm going to use. Combine these images. Place the vector logo from the second image onto the front of the white coffee mug from the first image. The logo should be curved realistically around the mug's surface and follow the studio lighting showing highlights and shadows consistent with the ceramic material. And I'm going to hit enter. Notice I'm asking for two specific things. Wrap it around the curve and match the lighting. Look at the result. This is what separates a basic edit from a professional one. See how the logo isn't just flat. It actually bends around the curve of the ceramic. And look at the lighting. There's a slight highlight on the right side of the logo matching the light source on the mug. It looks like a real printed product. You can use this to test your designs on t-shirts, tote bags, or packaging in seconds without needing a professional photographer. Before we wrap up today,

Conclusion & Final Tips

I have two quick power tips. First, don't forget the aspect ratios. If you're making YouTube short, or Tik Tok, add aspect ratio 9 to6 to the end of your prompt. If you want a wallpaper, use 16 to9. Second, if you have access to Gemini events, try toggling on thinking mode if your prompt is really complex. It takes a few seconds longer, but it reasons through the prompt before drawing, which often fixes issue with hands or complex scenes. I've put a list to all the prompts I use today in the description below, so you can copy and paste them to and try them out yourself. Let me know in the comments. Are you going to use this for logos, characters, or just for fun? Thanks for watching this time on Teachers Tech. I'll see you next week with more tech tips and tutorials.

Ещё от Teacher's Tech

Microsoft Excel Tutorial - Beginners Level 1

Teacher's Tech | 19.03.2018 | 21 сегм. | 13 463 596

How to Use Google Meet - Detailed Tutorial

Teacher's Tech | 24.03.2020 | 3 сегм. | 2 766 584

Microsoft Word Tutorial - Beginners Level 1

Teacher's Tech | 12.03.2018 | 26 сегм. | 2 626 025

Microsoft Excel Tutorial - Beginners Level 2

Teacher's Tech | 06.08.2019 | 10 сегм. | 2 614 553

How to Use Google Classroom - Tutorial for Beginners

Teacher's Tech | 30.03.2020 | 13 сегм. | 2 582 540

Vlookup Function in Excel for Beginners

Teacher's Tech | 24.02.2020 | 5 сегм. | 2 073 225

Ctrl+V

Экстракт Знаний в Telegram

Транскрипты, идеи, методички — всё самое полезное из лучших YouTube-каналов.

Подписаться