Exploring creative limits between man & machine

I’ve officially kicked things into full gear for this month-long project.

In case you’re wondering what this is, feel free to read this for full context.

TLDR; I’m creating a character daily for October. The goal is to explore creative boundaries alongside the usage of AI, by using my own drawings and without referring to any particular art style.

Note to self:

• Creating moodboards on Pinterest was a good idea

• I’m conflicted about some of the character designs due to different art styles

• So far, hoods, jewellery, masks and tech are a recurring theme

• I’m currently leading with a visual-first approach, perhaps a storytelling approach could improve things

• Greenkeepers, Council of Cats and Brotherhood of Owls are most exciting to expand on since they’re all factions

• On an individual level, Sentient Shroom, Five Fingers and Witch + Wendigo have the best backstories in my head – all tragic, haha

• I like Rubbit the Frog, Water Lord and Manablade Seth for the character design and abilities

One of the things I could work on is preparing my list of deliverables in advance, this would add time for inspiration.

Post-drawing, it would be interesting to share the origins of the idea and how I add my twist to it.

My sketches for the first 12 character designs

My 4-step process

1. Sketch my character concept

2. Translate the idea into a word prompt

3. Generate visuals with AI (this round, it’s with Stable Diffusion)

4. Analyse variants and note down shortcomings for future iterations

Here are my Stable Diffusion renders:

Stable Diffusion’s interpretation of my work based on text generated by ChatGPT + my notes

As a reminder, I’ve also written down my thoughts about its design accuracy and improvements worth implementing. Here are a few things it’s good at:

• The obvious: speed and variations. We are outmatched here, it has an arsenal of artistic styles, and options within clicks for multiple formats

• Being proactive with different design choices, bending the rules of given word prompts by referencing its database of a specific topic

• Ability to exploring finer details of a character without request (eg: weapon, accessories, etc)

Shortcomings:

• Staying neutral to racial stereotypes, it whitewashed one of my characters (dark skin + blonde hair) lol

• Creating functional layers that separate subjects from each other (this will cause problems for real world applications)

• Fingers and weapon placements are still a hit or miss, often distorted because of uncertainty

• Texts and notes written on beside the sketch/drawings almost always came out illegible, so I’ve removed them entirely