Text-to-image artificial intelligence (AI) tools that generate art (i.e. DALL-E 2, Midjourney, Stable Diffusion, Nightcafe, Artbreeder, etc.) have eliminated the artist’s block of the blank canvas for everyone.
Now, anyone can make masterfully rendered images by typing a couple of words into a computer or by uploading an image.
AI art tools have supersized creativity in the art world by the algorithmic manipulation of pixels to quickly replace humans for the fast cation of rough artistic concepts.
Even as some creative roles may be replaced, AI art tools are also opening avenues for new AI artists to evolve.
Yet, the more fantastical, chimeric imagery I see of people prompting in the style of established artists the AI-art platforms have scraped to fill their image data bases, the more I want to see how the average prompter would finish an image with an art style of their own.
I prefer to view images that text-to-image AI tools render as an additional reference for my own expression. I shared this sentiment in another article I wrote about using generative AI as an artistic collaborator.
Since the area of AI-generated art is developing at quantum speed, I thought I would provide an updated example of how I used a text generated AI image as a starting point to stylize with my own art.
I hope this example offers ideas to inspire you to try personalizing your own AI-generated images.
Start with your favorite text-to-image platform
I won’t go into a listicle to review all the various AI art generators that are available because there are many.
My recommendation is to do a little research to see what best fits your needs and your wallet. AI art generators vary in their learning curve and cost.
Here are some resources I found that might help get you started: Domestika.org, AI Artists.org and jonathandinu/awesome-ai-art
Conceptualizing with a prompt
For my example, I wanted to see if I could get a good concept image to visualize a dream. This dream involved luscious water flowing through ancient ruins. It was a cool dream.
I decided on the text description: beautiful opalescent waterfall with ruins
I chose three different AI art platforms to test my prompt, Midjourney, DALL-E 2 and Nightcafe.
As you can see by the images, each AI platform provided nice outputs from the prompt. Here are my thoughts on each interface.
Midjourney — I prefer to use Midjourney for my concept creation because of the artfulness of the outputs. Prompt design is a little more involved to get a specific result. Users must have a Discord account and the first 25 images are free.
DALL-E 2 — I haven’t used this platform as much as it only recently came of beta. Outputs are more photorealistic. Simplistic web-based plug and play user interface. Users are given 15 free credits to start creating.
Nightcafe — This platform has the most algorithmic options to choose from to create images. Because it has so many options, the user interface can be confusing with a lot of pop-ups. Users get free credits to start. E-commerce is built into the sight to purchase image outputs.
As I mentioned, all three images hit the mark as good concepts from my prompt. The composition of the Midjourney image is by far the best for drawing the eye around the scene and its waterfall is the most mesmerizing.
I moved forward with the Midjourney image.
Stylizing with neural style transfer
Beautiful as it may be, the Midjourney image is not my style. In addition to my digital design work, I paint, draw and sew. None of my creations have ever been in the style of dark fantasy.
I could have written more prompts to evolve the Midjourney image to fit my style, but I decided to transfer my style directly using Adobe Photoshop’s updated neural style transfer filter.
Neural style transfer imparts the style from one image to the content of another image.
Photoshop has improved their neural style transfer filter to transfer style with less content loss of the original image. Also, there is an option to upload your own image.
If you don’t have Adobe’s Creative Cloud of Apps and are code savvy, here is a link to set up neural style transfer on your local computer.
I picked three of my own art works to try the style transfer process on the Midjourney waterfall image.
Here are my results:
Once again, I liked all three images, but Style transfer 1 created the brightness, texture, and the color palate I was seeking.
Dark fantasy converted to JEFS painterly.
Adding that extra finish to make it really yours
What is the purpose? I ask this question for everything I create.
If I needed a quick way to create a great concept image, I would plug and play with a text-to-image AI-tool and be done. I can’t create anything as quickly as the AI tools.
The purpose of this project was to recreate my dream into a picture with AI tool collaborators. I make a work really my own with how I add finishes.
I liked the Style transfer 1 image, but the style transfer removed the depth of the lights and darks that make the Midjourney image so appealing.
To add back depth to the Style transfer 1 image, I used Photoshop to colorize the sky and water. I also adjusted the brightness and darkness of the ruins and waterfall sections.
Any photo editing software could be used, but I like to squeeze the worth out of my Adobe CC subscription.
I decided the scene also needed an observer and a creative statement.
To finish, finish the image. I used Illustrator to add a character I created, to texturize and to include my image statement.
The article cover image is the completed project.
Final thoughts
Advancements in AI-art tools have revolutionized creative expression.
While some people may fear the evolution of algorithmic art, the development of AI tools has helped me understand more about who I am as an artist and what I want to say.
Yes, now everyone can render amazing things with a click of a button.
I shared this example to show how text-to-image AI images can be used as seeds to add one’s own personal style.
Art is something that expresses important ideas or feelings. Make sure that what you are creating expresses who you are.
Thank you for any feedback or ways you may have found to use AI to inspire your art.
Like this story? Contact JEFS or sign up for news and updates
Julia Fletcher is founder of JEFS Storytelling Arts, a graphic design studio, where she uses her unique research skills and artistic talents to create custom visual stories that help clients’ increase engagement and promote the education of their audience.
Comentários