Stable Diffusion is a powerful AI system that can generate images from text descriptions. It uses a technique called “diffusion models” which starts with random noise and gradually refines it into an image.
Dreambooth is a method to customize or “fine-tune” Stable Diffusion models on new datasets with very little data. Typically only 3-5 images are needed to specialize a model on a new topic.
Together, Stable Diffusion and Dreambooth allow anyone to easily create AI systems tailored to their unique needs. In this article, I will share prompt examples and tips to get the most out of these tools.
What You Can Do
With Stable Diffusion and Dreambooth, you can create AI systems to generate:
- Photorealistic images of yourself, pets, objects etc.
- Art in any style – oil paintings, anime, pixel art and more
- Product photos – generate infinite product variations
- Fashion model photos – customize poses, styles and more
The possibilities are endless! Let’s look at some prompt examples.
Prompt Engineering Fundamentals
Constructing the perfect prompt is key to getting great results with Stable Diffusion. Here are some fundamentals:
Describe the Subject in Detail
Provide lots of descriptive details about what you want to generate:
A photorealistic image of an elderly man with short white hair and beard, wearing small round glasses, standing in front of a blue backgroundSet the Style
Specify an art style or medium:
An oil painting of a bowl of fruit including bananas, apples, and orangesGive Context
Add relevant context and setting details:
A fashion model wearing a pink ball gown poses elegantly on a grand staircaseDirect the Composition
Guide the position, framing and lighting:
A symmetrical close up portrait photo of a happy corgi looking directly at the camera with a shallow depth of fieldNow let’s see some full prompt examples.
Stable Diffusion Prompt Examples
Here are some prompts to showcase what’s possible with Stable Diffusion:
Photorealistic Portraits
A photorealistic image of a 20 year old man with short curly brown hair, brown eyes, light beard, wearing a navy blue shirt, looking happy while laughing with his mouth openThis produces a realistic generated image of a laughing young man.
Stylized Character Art
A dynamic digital painting of a strong female warrior with long red hair wielding a sword and shield, intricate fantasy armor, dramatic lighting, highly detailed, digital painting by artstation trending artist Greg Rutkowski and Alphonse Mucha This prompt generates a fictional warrior woman in a painted style. Specifying multiple artists helps blend styles.
Product Photos
A photograph of a new smartwatch with a round silver metal case, rectangular black touchscreen, and blue silicone watch band, resting on a wooden table with soft lightingThis prompt clearly describes a smartwatch to produce a realistic product shot.
Tips for Better Prompts
Follow these tips when writing prompts:
- Be specific – Provide lots of descriptive details
- Set the style – Specify an art medium or style
- Give context – Add relevant context and setting
- Direct the composition – Guide positioning, framing and lighting
- Specify multiple styles – Blend styles by listing multiple artists
Spend time crafting the perfect prompt. It makes all the difference!
Dreambooth Prompt Examples
Dreambooth allows training AI models on new topics with very little data. Here are some examples for training custom models:
Pet Model
Images: 5 photos of your pet from different angles and backgrounds
Prompt: A photo of my pet corgi named Biscuit sitting on the couch looking happy
This trains a model to generate images of your pet.
Person Model
Images: 3 portraits of yourself with different poses and expressions
Prompt: A photo of myself wearing sunglasses and a blue shirt, making a silly face
This creates a personalized model to generate unlimited new photos of yourself.
Product Model
Images: 4 product photos from different angles on white background
Prompt: A photograph of the new Acme wireless headphones resting on a table
This trains a model to render product images.
As you can see, Dreambooth makes it easy to create AI systems tailored to your needs with very little data.
Helpful Resources
Here are some helpful sites for learning more about Stable Diffusion and Dreambooth:
- Automatic1111 Stable Diffusion WebUI – Web interface for easily using Stable Diffusion
- Stable Diffusion Notebook – Jupyter notebook for running Stable Diffusion locally
- Dreambooth Notebook – Notebook for training Dreambooth models
- Lexica – Discover and share prompts
- PromptHero – Find the best prompts
So in summary, Stable Diffusion offers powerful AI image generation from text prompts, while Dreambooth makes it easy to customize models for your unique needs.
I hope these prompt examples and tips help you create your own AI systems! Let me know if you have any other questions.