Doraemon’s AI Gadget: Create & Transform Images with Magic!

Nobitaaaa! 😃 Have you ever wanted to create amazing images just by saying some magic words? Well, guess what? With Stable Diffusion, you can! ✨ Whether you want to create an image from scratch or modify an existing one, this AI tool is like one of my secret gadgets from the future!

Today, I’ll show you how to do two amazing things with it:

1. Text-to-Image Generation — Just describe what you want, and voila! An image appears! 🎭
2. Image-to-Image Transformation — Take an existing image and tweak it into something even cooler! 💡

And the best part? You can do it all on your own computer with just 6GB of GPU power! Ready? Let’s dive into my futuristic AI toolkit! 🛠️

1️⃣ Poof! Creating Images from Text (Text-to-Image)

Gian once asked me, “Doraemon, can you create a cool image without using a camera?” And I said, “Of course!” With Stable Diffusion, all you need is a text description, and it will generate a masterpiece! 🖌️

First, Install the Magic (Dependencies!)

Before we can use this futuristic tool, let’s install some necessary packages. Run this command in your terminal:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
pip install diffusers transformers accelerate safetensors

💻 Now, Let’s Create Some AI Magic!

import torch
from diffusers import StableDiffusionPipeline
# Load my secret AI gadget! 🤖

model_id = "stabilityai/stable-diffusion-2-1-base"
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe.to("cuda")  # Zoom! Moves to GPU! 🚀
# Say the magic words! ✨
prompt = "a painting of a cat"
image = pipe(prompt).images[0]
# Save the creation for Nobita to see! 🎨
image.save("generated_image.png")
print("Image saved successfully! Yatta! 🎉")

Output

What’s Happening Here?

• Loads the Stable Diffusion 2.1 Base (Super-efficient!)
• Uses torch.float16 to save GPU memory ⚡
• Generates an image based on the text (“a painting of a cat”)
• Saves the result so you can show it off to Shizuka-chan! 😉

But wait, there’s more! Want to edit an image instead of creating one from scratch? Keep scrolling!

2️⃣ Transforming an Image with AI Magic (Image-to-Image)

“Doraemon! I want to change this cat into a Persian blue cat!” — That’s what Nobita asked me. Well, no worries! With the img2img pipeline, we can modify an image while keeping its overall structure intact! 🎨

💻 Let’s Get Started!

import torch
from diffusers import StableDiffusionImg2ImgPipeline
from PIL import Image

# Load the gadget again! 🛠️
model_id = "runwayml/stable-diffusion-v1-5"
pipe = StableDiffusionImg2ImgPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe.to("cuda")  # Boom! GPU activated! 💥
# Load the generated image from before 🔄
input_image = Image.open("generated_image.png").convert("RGB")
# The new transformation spell! 🎭
prompt = "change the color of cat and breed to persian blue cat with white fur"
# Poof! Transform the image!
output_image = pipe(prompt=prompt, image=input_image, strength=0.6, guidance_scale=7.5).images[0]
# Save for Nobita! 🖼️
output_image.save("output.jpg")
print("Image transformation complete! Whoa! 🎉")

Output

🧐 What’s Going On Here?

• Strength = 0.6 — Controls how much of the original image is kept. (Higher = more changes!)
• Guidance Scale = 7.5 — Tells the model to follow the prompt more strictly! 📜
• Original Image → New Image — We just changed a regular cat into a Persian blue cat with white fur! 🐱✨

Optimizing for a 6GB GPU

My dear Nobita, not all gadgets are built the same! Here are some tricks to run Stable Diffusion smoothly on a 6GB GPU:

• Use torch.float16 to reduce memory consumption. ⚡
• Use a lower strength value (e.g., strength=0.4) to process images more efficiently.
• Stick to Stable Diffusion 1.5 or 2.1 Base (because SDXL needs 12GB+ VRAM). 😲

Conclusion

Wow! With just a few lines of code, we:

✅ Created brand-new images from text prompts! 🎭

✅ Modified existing images using AI magic! 🖌️

✅ Did it all on a 6GB GPU without breaking a sweat! 🏋️‍♂️

Stable Diffusion is like a gadget from the future, bringing AI-powered creativity to everyone! So, whether you’re an artist, developer, or just someone who loves cool technology, go ahead and experiment! 🚀

Nobita, should we turn this into a cool GUI or an API next? Let me know in the comments! Hehe! 😆🎨