Production Update: It's full steam ahead on wrapping up the music composition! We'll be moving to full-time production mode starting November 1st.
I'll post updates to the Ko-Fi page, stay tuned! |
|
|
Waking Up AI Art Contest: Results
Back in July, in a fit of COVID fever, I decided to organize an online art contest to come up with the cover art for Waking Up, and since the relationship between human and machine is so central to the story, I figured I'd make it AI only -- you're only allowed to submit works where AI made the final decision for every pixel.
Out of 60-some applicants, I chose seven. I paid them a stipend for their time and to reimburse AI tool subscription expenses. I gave them a month and chewed their ear off about the story, really I can be longwinded when allowed. We checked in weekly where I asked for detailed feedback on the process -- with mixed success in availability I have to say. I suppose we're all adults with other commitments. Anyway, I got five submissions back, and loads of great write up on different perspectives and techniques. I'll be compiling those and making them available on the website soon... or maybe the paid section of the Ko-Fi page? That's how this stuff works right?
Anyway here are the results: |
|
Winner: Joanna Gonzalez
On her approach:
I approached the concept by reading the Waking Up script and also referred back to the producer's summary write up. The script allowed me to gain a more detailed and nuanced picture of the plot and character and the producer's summary gave me a more thematic and bird's-eye view of what Waking Up wanted to explore at its core. I retrieved insights from the summary like visualizing consciousness as a vibration, described in the Tibetan Book of the Dead, which I wouldn't have quite gotten from the script alone. It ended up being a central theme I focused on for the design visuals and word phrasing I incorporated in the prompts as I moved forward with the concept.
On her process:
There were several times I was mostly happy with a design but would have liked one or two elements altered, and although there are workarounds for this, the results weren't quite the same, which still produced very interesting and curious images but at times went outside the scope of what I envisioned. There was a particular prompt I started getting the best results with so I continued generating multiple versions, remastering them, upscaling them, and just continuing a queue of chains that eventually evolved into the final image. I also ended up simplifying my prompts instead of crowding them with too many keywords and commands. I started to realize, at least in my experience, that the more focused, concise, and clear the details were in the prompts, the more compelling the results were.
On using the tech moving forward:
I would definitely continue using AI art tools as they continue to take shape for my future projects and also to conceptualize creative directions and ideas. I learned how different of a medium these AI tools are for the imagination and the lessons in language they unexpectedly provided ranging from how to more thoughtfully construct a prompt with phrasing to having a more focused understanding when describing art movements, styling, texture, lighting, perspective, context, and details that may be overlooked when creating a composition. But even with a seemingly comprehensible arsenal of keywords, the liberties the tools take are unexpected and provide an element of surprise.
|
|
 |
I can't say enough how much I love this submission, nor how hard it was for me to choose an ultimate winner, as each contestant's journey took them in a completely different direction, not to mention being able to appreciate each perspective. There was an element of relevance to the story in the scoring, and ultimately I went with the one that I thought would attract the most prospective listeners and not confound their expectations.
Please consider chipping in to cover Joanna's prize money. Just click this link and then the purple donate button.
I really liked every submission that came in, so I figured I'd share them with you. You may also appreciate some of the creative tricks artists used to push the boundaries of "AI-generated art." |
|
|
 |
Monica Trego
"...the prompt that generated this one on Midjourney was "consciousness as a vibration". In reading the brief I found that an interesting concept, and I kept coming back to it. I worded it in multiple ways, with different styles applied, but in the final week I used that exact term with no style cues...and my image came to life. Both concept and image are simple, ethereal and thought provoking." |
 |
Vinh Nguyen
"...I'm choosing this one partly because of the more outwardly "digital" feel, but it's a close decision between a few variants. This comes from my last push off this triptych that veered, clearly, and went into more metaphor than explicit request.
I'm also choosing this because out of all of my generations and variations, this last track of ideas read best in both thumbnail and larger formats." |
|
 |
Delta Abbott
"...After reading the script, there was one scene in particular that inspired an image in my mind. I took this idea to Midjourney to experiment but it was difficult to create the scene in my head. After many variations and upscales in Midjourney, I settled on an image which I brightened in photoshop, then moved to DALL-E to edit. I outpainted the right side of the image to reframe it and used the inpainting tool to add the main character and other details to align with the script. Once everything was where I wanted it, I used the image editor in Stable Diffusion to smoothen details and give it a more cohesive feel.
I've really enjoyed being a part of the project and having a chance to experiment in different AI art generators, I'll likely use them in future to come up with concept art and ideas or inspiration for my own work." |
 |
Ritvik Takkar
"Adding things like octane render, volumetric lighting really helps to get a desired outcome also I think writing detailed prompt is yielding better results than short prompts." |
|
|
|
|
