Exploring and learning
I've been fascinated by the rapid advancements in Generative AI technology and the incredible potential it holds for businesses and creative professionals. This project documents my journey learning about, experimenting with, and researching various generative AI models and applications.
What started as curiosity about text-to-image generators quickly evolved into a deep dive into the technical frameworks, terminology, and inner workings of generative AI systems. This documentation traces my learning journey.
Research Framework
Implementation Framework
Workflow Progression
Diffusion Process
Exploration Objectives
When I began this journey, I established clear objectives to guide my exploration of generative AI technologies:
Understand How Generative AI Works
Explore the mechanisms of diffusion models and transformers
Master Terminologies
Learn the vocabulary to effectively use generative AI tools
Gain Creative Control
Develop techniques for precise, controllable image generation
How Generative AI Works: Text to Image
Understanding the fundamental process of how text prompts are transformed into images.
Text Encoding
Prompt encoded to numeric representation using models like CLIP
Diffusion Process
Model refines random noise into images through denoising steps
Sampling Methods
Algorithms (Euler, DPM, etc.) control noise removal
Common Terminologies in Generative AI
Terms that appear often when using image generation software.
Checkpoint (ckpt)
Saved model state specialized for different styles
CLIP
Translates text into embeddings that guide image generation
VAE
Compresses images into latent space and decodes back
Latent Space
Compressed representation where diffusion occurs
How ControlNet Works
ControlNet adds precise control over generation using conditioning inputs like edges, depth, and poses.
Different Models and Their Applications
GANs
Transformers
VAEs
ComfyUI Workflow
Base Model Output
Advanced Technique Output
Design Application
Marketing Application
Content Creation
Prototyping Example
Thoughts on the Future of Generative AI
After deep diving into the rabbit hole of Generative AI, the possibilities seem endless. Anyone can create beautiful images, songs, and art with these powerful tools.
Technical Fluency
Understand AI terminology and concepts to direct tools effectively
Curation & Refinement
Critically evaluate and refine AI outputs
Strategic Integration
Know when and how to integrate AI in the process
Human Connection
Focus on nuanced human needs AI might miss