Unleashing Your Creativity
A Beginner's Guide to AI Art in 2023
Welcome to the exciting world of AI art. It's a realm where technology and creativity collide, and the results are nothing short of mesmerising. But what exactly is AI art? Simply put, it's any artwork created with the help of artificial intelligence. Whether it's an image created from a short sentence or an animation of your most recent selfie, it takes dedication and skill to be able to create something truly remarkable.
The main current question on many lips is, “Is AI art a science or an actual art?”. Whilst there is a science behind the most effective way to ‘prompt’ or communicate with the AI tool of your choice, AI-generated art can be highly detailed as a result of the creativity of the artist. True AI artists can intricately manipulate and enhance their creations with high precision. It takes dedication and skill. Whilst there is certainly a science to it, if that's not art, I don't know what is.
It can understand and interpret sentences that include multiple clauses, and generate coherent images that combine all the different elements into a semantically cohesive whole. This allows users to give detailed instructions on the type of image they want to create and get a high-quality result.
To choose which platform best suits your needs, you have to experiment. There is no other way.
For those curious about how AI Art platforms are built, without getting too technical, it's all about the training. Most of the popular platforms (not all) that create AI art are based on the technique called Generative Adversarial Networks (GANs) to teach their engine to create new images that are similar to the ones in the dataset. We won’t get into GANs today although you can click 👉 for Wikipedia’s entry.
Tools, Platforms & Prompt Engineering:
The rate of computational processing power is growing rapidly. We’re on an exponential growth curve unlike anything seen before. As a consequence, there are a plethora of tools and platforms to help you generate AI art. Many are built from open-source code (publicly available foundations), some are closed-source (foundations are not publicly available) but all are currently vastly different in terms of their user experience and interface. The foundations of these tools are tweaked to the company’s preference and each have different data sets (they are trained off different images) so you will get varying results depending on the platform you use. The exponential increase also means the tools will only get better. Speed to generate the image will quicken and the ability to create exactly what you have in mind will become perpetually easier.
Interaction with each platform varies although, most of the time, you will need to include the ‘prompt’, which is a description of the image you have in your head. There are other platforms in which you can generate a similar image from an existing image although we’ll be focussing on “prompt engineering” so you can create something truly unique.
We’ll only focus on three tools today. They all produce different results/outputs so will give you a good understanding of the different capabilities.
Prompt Engineering:
Computers are stupid. Miraculous, but stupid (future AGI - please don’t turn me into a paperclip for this one!). You need to tell a computer exactly what to do in order for it to do it. The elegance of a simple interface like Google Search only works because humans told it exactly how to work. The same goes for AI Art. The more articulate you are, the more time you spend creating your prompt, the better your image “output” will be. Hence, you need to ‘engineer’ your prompt.
Prompt engineering is the process of designing and creating the initial text or ‘prompt’ that is used to generate the image. The prompt sets the context and provides the model with the information it needs to generate a relevant and appropriate output. The goal of prompt engineering is to create a clear and specific prompt to guide the model's output while also being open-ended enough to allow for creativity and variation. Prompt engineering can be used to generate a wide variety of content, including text, images, and audio.
By offering a collection of key phrases that serve as guiding prompts, you can generate various outputs that align with the specific style you have selected. From artistic forms and designs to legendary artists and genres, keywords and sub-categories represent the various styles you can choose from. Lighting, rendering, chaos and more are also at your disposal. Your options are literally bound only by human creativity:
Tools:
DALL·E 2
DALL·E 2 is an AI system created by OpenAI that is capable of generating and editing images based on natural language instructions. Want an avocado armchair? Consider it done. A dinosaur wearing a sombrero? Easy peasy.
DALL·E 2 is also able to expand images beyond what's in the original canvas. This is called ‘Outpainting’. It can create expansive new compositions by adding and removing elements, while taking shadows, reflections, and textures into account. This allows users to create unique images that are not limited by the original image's size or composition.
It can add and remove elements, change the size of objects, and even change the background of an image. It can also understand the context of the image and make changes that are consistent with the scene. This is called ‘Inpainting’.
MidJourney
Like DALLE-2, MidJourney allows users to create AI-generated artwork by simply entering a text-based prompt that describes the image. There are a few key points of differentiation compared to DALLE-2:
It runs (almost) entirely within the Discord platform, which makes it more accessible to users (a webapp has just been released although you can use it for free via Discord). Its training data was mostly creative art so if you want photorealistic images, you may want to use a different tool. Realistic outputs can be achieved although it’s not their forte.
The noise of Discord can be easily overwhelming so if you don’t want to pay for the premium version and want some helpful tips, check out the YouTube tutorial below. It provides some key pointers to make your Discord/MidJourney a bit more fun.
MidJourney V4 is fantastic at small detailings in all situations. It can handle complex prompts that contain multiple details, and it is better with multi-object/multi-character scenes. It also seems to have an increased understanding of creatures, places, and environments.
Upscaling and creating different versions of your creation are easily done. Watch the below video for step-by-step instructions.
Lexica
Lexica.art allows you to search existing images made via its platform and also create images. The platform is easy to use, it has a search bar that users can use to search for specific prompts or prompt elements. In other words, you can see the image that was created by another user and the prompt they used. Making it easy to find similar artwork to what you have in mind. The results/outputs are generated quickly.
Some of the key features of Lexica.art include:
Ability to choose from V1 or V2 models.
Trained on a variety of images ranging from art to high-quality photography.
Upload a picture to see similar user-generated outputs (can be done from your smart phone)
Ability to change the “Temperature/Guidance scale” - allowing for more AI creativity or an output that is closer to your exact prompt.
Variety of sizes: The platform allows users to create different sized images, including landscapes, portraits, square and other dimensions.
Outpainting
Showcase
Twitter is home to a vibrant and dynamic community of AI artists, who are pushing the boundaries of what is possible with this cutting-edge technology. From the mind-bending creations of @DocT___, to the ethereal landscapes of @HODLFrance, these artists are dedicated to exploring the full potential of AI as a medium for creative expression. Their work is a testament to the boundless creativity and imagination that is possible when humans and machines collaborate to create something truly unique and beautiful. So, if you're looking for inspiration, or simply want to experience the latest in AI art, be sure to check out these amazing artists on Twitter and see for yourself what the future of art looks like.
Animation Experts
Conclusion
Please bear in mind that this post is only skimming the surface of what’s possible. You are bound only by your imagination. The possibilities are endless.
The artists mentioned above are also just a small sample of a vast collection of extremely talented artists. You’ll see that once you get started generating your own images, there are many other layers to create such a standard.
The platforms and tools available to artists today are more advanced and user-friendly than ever before. And with the work of talented artists like the above, we can see just how far this medium has come. But the true beauty of AI-generated art lies in its ability to constantly surprise and inspire us. With each new creation, we are reminded of the boundless potential of technology and the human imagination. AI Artists spend hours, days and weeks perfecting and crafting the image they have in their heads. It is an intricate and creative process that requires thought, precision, trial and error. If that isn’t art, I don’t know what is.
If you enjoy Artificial Intelligence discoveries, please consider joining my weekly newsletter. I break down several resources to help you on your AI journey.
Onwards, Deiniol
