What is AI Imaging?
In a nutshell, AI Imaging can be defined as the science (and art) of using Artificial Intelligence technology to generate, enhance, edit or replicate images. This can be achieved using text, speech, or existing images.
Our main focus going forward will be on Text to Image technology, which is probably the most popular and widely used form of AI driven Image generation.
#Text to Image for creating images
The concept of generating images using text to Images is pretty simple. One has can enter a command, or a prompt, which defines what type of image one wants to be generated. Based on this prompt, the AI engine used by a particular site or app will start its work in the background and generate that image for you.
This very simple premise, and a very interesting concept, which uses complex algorithms, deep layers of technology, and quite a bit of expensive hardware (particularly graphic cards) to produce amazing images.
AI Imaging offers several opportunities, such as
- Creating different versions of the same image
- Generating multiple images in a very short span of time
- Digital images are often optimized for size (most images created using Stable Diffusion 2 are smaller than 100 kilobytes in size)
- Provides an opportunity for 'trial and error' : with multiple art forms, materials, locales, cultures, and so on.
- Many images have a unique Identity number- called 'seed' that can be used to create variations or copies of a previously generated images.
The technology is not without its limitations. The most discussed and/ or the most frustrating limitation is that very often, the limbs of the human subjects are missing, or are deformed. Other limitations include generation of NSFW images- though many sites have anti nudity, violence or profanity filters. We will discuss this particular aspect in a subsequent chapter. Also note that the resolution of the images is often on the lower side- typically 500x500 square or less than 1000 pixels in landscape or portrait mode. This is because of the high amount of computing horsepower required for generating the images. This limitation may become less relevant in future versions of the imaging technologies.
#Some Points to note
Before we dive further, I take this opportunity to highlight or clarify a few things. Let us call them housekeeping rules.
First of all, the focus of this book will be on AI Image generation using Text to Image or Image to Image tools. We will also cover the areas of enhancing images (i.e. increasing the resolution, improving image quality, etc). Other forms of AI content creation such as text or blog writing, AI generated music or videos are not covered in this book at this time.
Secondly, I consider myself to be a novice in the field of graphic design or art in general. This book is intended for people like me : the tinkerers, the explorers, and the novices to AI generated art.
As a side note, I have included an entire chapter (Chapter IX) which includes the glossary of terms and definitions that will help you understand what AI imaging is all about.
Dogs Galore- video collage of images of Dogs created by me, images rendered by AI.