Table of contents
- The beginning of an era of co-creation between humans and machines
- Explanation of co-creation process.
Two weeks ago, like many others, I finally got access to the Midjourney Beta after months of waiting. Developers keep sending out new invitations, so chances are you’ve heard them or seen them shared on social media. I’ve been enjoying myself in the generative arts community since 2018 and have had my fair share of tools powered by machine learning.
Recently, the wave of artificial intelligence has surged, and there has been a rapid increase in attractive art generated by AI. What sets Midjourney apart from other [generative AI models] is how fast it works. Results that used to take minutes or hours can now be produced in seconds.
WOMBO’s iPhone app “Dream” is the only tool I can think of with similar speed, but Midjourney’s results are incomparably better. Of course, I must also mention other alternatives such as Google’s Imagen , Disco Diffusion , or Open AI’s eponymous Dall-E (By the way, an invitation for anyone reading this to try Dall-E 2 . If you have received it, I would greatly appreciate it if you could share it..
AI-driven art may be a new field that creates works of art with words instead of handwriting. An artist’s talent in this field is measured not by his drawing or drawing skills, but by how well he communicates his vision through prompts.
The AI-driven art user’s toolset no longer consists of brushes and pencils, but fields occupied by vocabulary and glossaries. How an artist expresses complex ideas determines how accurately a program can interpret their intentions.
AI-driven art poses very important questions about the future of artists and their place in the creative process. Some think that concept artists, illustrators and painters are doomed and that AI (like LaMDA) will soon develop consciousness and take their jobs, but I think about the future of art. have a very different vision.
My view is that AI-driven art is driven by human-computer collaboration, and that in this form of art the artist becomes the art director. If you’re a photographer, graphic designer, or CG artist, you know how many days it takes in the early stages of a project, or in the ideation stage.
Experiments are usually expensive. Many times, as an artist, you’ve probably dreamed of having more time to practice and come up with ideas before entering the production phase. The research, references, and influences behind the final product will often remain invisible to the client or end-user.
How shocked he must have been when those involved were not satisfied and were forced to return to a blank slate. Often times, you’re forced to step out of your comfort zone and attend meetings when you don’t have enough time to develop fresh ideas. Tools of this kind (like Midjourney) can significantly reduce this ideation window by providing thousands of options in hours instead of weeks.
It also does a good job of A/B testing with different outputs such as color palettes, lighting scenarios, materials, textures and composition settings. Artists spend much less time building worlds and exploring characters, allowing them to invest in parts of the creative process that often go unnoticed.
Working with AI is nothing new for humans. For example, we are already familiar with AI through messages and emails, autocomplete functions in word processors, and advanced features in image and video editing software.
From voice assistants in mobile phones to enemies and NPCs in video games, AI has become a part of our daily lives. I am convinced that AI will be used in all industries in a few years, so learning to collaborate with it early is a good idea.
The beginning of an era of co-creation between humans and machines
Now back to Midjourney. The key to using this AI is interacting through Discord’s chatbot, and its prompt system is pretty powerful. It’s still in closed beta (as of June 2022), so I can’t show you the user interface, but I’m kidding, of course I will.
Before we do that, let’s define some simple concepts. You can add words or adjectives to the prompt, variables like artist name or art style, or parameters like aspect ratio or camera lens, and each additional word can dramatically change the final output. For example, the image below was generated entirely by Midjourney’s AI.
Let’s break down the prompt we used to generate the above image. A simple prompt can consist of one word, for example “Woman”.
But if you say “Black Woman,” the results would be very different. Most AIs, in particular, are trained with biased models and are not necessarily inclusive.
And you can add adjectives to make it more descriptive, like “A Tall Black Woman.”
Then add silly details like “A tall black woman wearing a red dress made of fire.”
You can also influence the results by taking inspiration from a famous artist, a particular art style, or medium (painting like oil or watercolor). For example, “a tall black woman wearing a red dress made of fire in the style of Leonardo da Vinci.” The image generated by input is as follows.
You can also combine sentences like this: “a kid drawing of a tall black woman wearing a red dress made of fire in the style of leonardo da Vinci” vinci)”
You can also add parameters. For example , you can specify the aspect ratio of a landscape image by adding to the end of the prompt . There are many other parameters, but I won’t explain them all today.
Now imagine all the possibilities. That’s an exaggeration, let’s really think about it for a moment. One of the ideas I want to realize is to replace traditional picture frames with Web 3.0. Imagine if instead of a single piece of art on your wall, you could replace it with an empty digital screen using tools similar to Midjourney.
Instead of texting though, it connects the system to voice AI (the same way people today turn on the lights in their apartments by talking to Google home, Siri, or Alexa). When you wake up each day and speak to this screen, you can ask your frame to visualize a different idea each time.
I would like to make an installation like this in a museum and have visitors print out what they like so that they can become part of the exhibition themselves. He is also interested in what the collective intelligence of the visitors will produce.
There are many other combinations of my ideas above, and I encourage you to take a look at some of the amazing projects the generative community is working on right now (even outside of the NFT realm).
Another interesting concept is for writers, authors, or poets to generate visuals for each page or paragraph of their work. Imagine AI-powered art helping filmmakers get inspiration for their screenplays, or immerse kids in reading their favorite novels.
All right… let’s take a deep breath here. Can you still keep up? I want to go even deeper into AI-driven art. I want you to hold on tight. Let me show you my normal process of co-creating with an AI called Midjourney.
Explanation of co-creation process
To give you an idea of how complex co-creation with Midjourney is, I’ve provided a map of my exploration through co-creation. Because conversations with chatbots are so linear that it’s easy to get lost in their interfaces.
My preference is to export the output to a design tool such as Figma and use this larger canva to map out my options and combine together what I find interesting, It is to produce more detailed results. As a sneaker fan, I wanted to imagine Nike’s biomimetic shoe collection. Here are the final results based on those ideas.
It all started with a simple prompt like this: “Nike sneaker made from coral reef.”
Midjourney will always generate 4 different images from your original prompt.
Based on this result, I decided to upscale the second image (enlarged display while maintaining high resolution).
Then I created a variant (a different version) from the second of these new images.
In this generation result, I decided to upscale the third image.
At this point, I liked the result so much that I decided to modify the prompt “Nike sneakers made from fish bones” a bit. Inspired by the work of my good friend Bahbou Floro, an artist from Martinique (a French island in the Caribbean). His fishbone creations are an artistic universe, and he loves sneakers.
Upscale the fourth result (enter modified prompt).
This result was also very interesting, so I combined this result with the first one to create a new set of variations influenced by two different prompts.
Further variants were created from the initial results.
Then I upscaled the final image.
When I finally combined this last result with one of the results generated from the first prompt, the new result became even more interesting.
What if we upscale the second result one last time!
I am really happy with all the results obtained. The image listing below is a map of all the images I generated, including the ones I ended up not using.
Let’s take a look back at the road we actually walked.
If you’re interested, I’m sharing the results of my daily explorations over the past two weeks. I tried to visualize ideas related to the world I was building inspired by nature, Caribbean culture, and a mix of science fiction and fantasy. Below is a link to the Twitter thread that posted the collection of ideas.
By the way, I believe that an artist’s actual work should not end with such co-creation. The above is just the curation stage. Of course, it’s all fun, and I spend a lot of time on the Midjourney website and on Discord looking for inspiration in the work that other communities have done.
But what’s even more fun is taking the final selections into tools like Procreate and Blender to paint and 3D model new sneakers inspired by AI art. With tools like this out there, concept art will never be the same as it used to be. Again, I don’t think this tool will take people’s jobs…I’m sure it will expand. I am already addicted myself and will continue to use AI in my daily process to take my art to new heights.