Iss. #4 AI Uses for a Multi-Disciplinary Creative
The Machine and the Flower AI reference image

Iss. #4 AI Uses for a Multi-Disciplinary Creative

I truly love creating—composing songs, writing poems, illustrating worlds, developing characters, producing videos, telling stories. Creating is my biggest passion, and for me, my love of creating lies in the process. Its like building a puzzle—the art on the cover of the puzzle box isnt the point; it's the building of the puzzle that brings the puzzle-builder joy (and yes, at moments, anxiety and stress too).

So you can appreciate why AI will never replace the creative process for me. When it comes to my own creative works—my stories, my illustrations, my books, my poems—I want to be the one who creates them, because the process is the point, for me. (Also, those artists and professionals familiar with AI very well know that AI is not yet there when it comes to quality when compared to professional writers, artists, illustrators, etc.)

With that said, I am excited about some of the prospects and promises of AI. Like any tool, it is not good or bad—like with any technology or tool, there are opportunities, challenges, and dangers that it presents. As a technology that undoubtedly will have a huge impact on almost every aspect of life, it must be designed and developed with caution and intention.

I have a lot of thoughts about AI and machine learning, a lot of it philosophical, a lot of it having to do with how it will redefine the artist and the craft of an artist. But this article is more practical.

Though I've said that AI doesn't replace the creative process for me, I do use AI in various ways to support some of my different creative processes. I thought I would share a few here. Perhaps they will give you some new ideas or thoughts on how AI can assist your own creative processes.

1. Illustration-Create your own textures

One of my illustration techniques involves blending texture images or photos in with elements in my illustration. For example, for a character's shirt, I often like adding a picture of cloth and blending it with the shirt's solid color to give it more depth and, well, texture.

It's a pretty common technique, and traditionally an artist (like me) would either take their own photographs of different textures and/or use stock footage of texture images. With AI, you can literally create any texture you'd like with an image based AI tool like Dall-E.

Be prepared to tinker and make adjustments to your prompts. It can be frustrating at times getting the AI to get it right.

"Create an image of rope texture. Make it a pattern that repeats across whole image. Horizontal/landscape dimension. Realistic."

2. Illustration/Writing-Reference Images

Visual artists and illustrators often use images as references of particular objects or body poses. Writers can also use reference images to enhance their imagery and help them describe a scene, object, or material. I was writing a short story called The Machine and the Flower, which is about a flower who befriends a man-made war robot in a post-apocalyptic Earth setting. I generated a few images with AI of a rusty, moss covered robot, not to use for my illustrations, but to help me brainstorm descriptive words and phrases.

(Here is the story of The Machine and the Flower, if you'd like to read: https://medium.com/@mrpendlum/the-machine-and-the-flower-def24353c463 )

AI reference image for The Machine and the Flower

3. Creative Writing-Research and Accuracy Check

One of the poetry collections I'm currently working on is Tome Terra. It tells the story of life and earth chronologically, from the formation of the first stars to the beginning of cellular life, to the evolution of complex life, and beyond. From cyanobacteria to protists, Snowball Earth to the Cambrian Explosion. Unlike my fictional poems, these poems require a lot of research to make sure I get to facts as straight as possible.

AI can definitely assist with research. When asked to reference the web on a topic, ChatGPT now even provides links from the resources that it referenced. One thing that I've started doing with each of my poems is an accuracy check. Essentially, I ask the AI to check the poem and give me a level of accuracy, with some commentary to go along its accuracy assessment. I also use it to check the chronological accuracy of the poems in relation to each other and in the wider context of the collection.

A word of caution: never rely on AI alone for research. Not only can it hallucinate and make mistakes, but it also isn't up to date on the latest findings. Just as an example—I was able to add a poem on dark oxygen—oxygen generated without light or photosynthesis (a pretty significant finding for the story of life)—when I happened to come across an article on some relatively recent research findings from 2024. ChatGPT was not "aware" of this research until I pressed it further and got it to search specifically for Dark Oxygen on the web.

So, yes, AI is great to supplement research, but don't depend solely on it.

4. Game Design-Brainstorming and Refining Game Mechanics

Slowly, over the long-term, I've been creating and designing a couple of trading card game ideas. The game mechanics of one of my games have been pretty cemented for some time; one of the next main tasks was creating abilities or actions for the different character cards of the game based on the mechanics.

I put together a thorough enough description of my game mechanics, then fed it to AI and asked it to put the mechanics together in a "Rules Booklet" format. It did a pretty good job, and from there, I was able to refine the instructions further.

I then asked the AI to give me a list of 50 possible "abilities/actions" based on my game mechanics, and indeed 50 it spit out. Some of them weren't applicable, but some of them were useable; others helped me think of other actions or abilities that I hadn't thought of.

All in all, I would AI is a pretty great tool for brainstorming and bouncing off ideas.

5. Video Production for Social Media-Poetry Visualization

A more recent use that I've started experimenting with is creating video with AI for the purpose of social media. Again, when it comes to my creative productions, such as my animated music videos, I am involved with the creation of the visual assets and work closely with professional animators to create the final results.

But I think AI can be a good supplement for quicker turnaround, social media based content. In my case, I've been experimenting with some visualization videos of some of my sci-fi poems (from my Alt Atlas poetry book). Essentially, I create short video snippets visualizing the extraterrestrial world from my poem, to which I add my poem narration.

Image of one of my poetry visualization AI-generated videos

A few final tips:

Video Generation

For generating video with AI, what I have found to be the best pipeline for me is: first I generate a good AI based image through Dall-E (or similar image based AI), and then feed the image to an image-to-video engine (such as Runways Gen 3 Alpha model) to create a video based on the image. At least for getting specific results, I find it easier to generate an image through Dall-E (or similar AI) thats to my exact taste, and then get the video AI to give it motion, though again, the results can sometimes be wonky and bizarre.

Trying to generate a video from scratch thats to my taste directly through Sora (or some other video based AI) hasn't been as successful for me.

Prompting

Even if your prompts are written out in great detail and care, AI still gets it wrong often.

Add a flower on top of the robot.

AI says "Ok, I have added a flower on the robots head, as requested," either lying through its teeth to my face, or just completely unaware that it has, in fact, generated the robot without a flower on its head for the twentieth time.

Adding detail to your prompts help, but sometimes (and this can depend on the model and type of AI) adding too much detail can overwhelm it and have the reverse effect (like when using video-generating AI). Currently, AI prompting is more of an art than a science.

One thing I've tried to do when working with image or video based AI is first give a language based AI (LLM) a full description of what I need, and then ask it to generate the ideal prompt to feed the image or video based AI. Its still hit and miss, but sometimes it does give better results.

It goes without saying that the best thing is to experiment and see what works for you.


These are just a few creative uses for AI. I'm interested in hearing if you have others that you use! Let me know in the comments!

Yevheniia Nyzhnia

Detail-oriented Legal Billing Specialist | Continuous Performance Improvement at Alvarez Stauffer Bremer PLLC

1 周

I agree

要查看或添加评论,请登录

Mr. Pendlum的更多文章