How I Tested One Sonic-Inspired OC Across 3 Totally Different Worlds
A small prompt experiment taught me that the same character can feel completely different depending on lighting, background, and composition.
When I generate AI character art, I usually care about one thing more than anything else: consistency.
It is easy to create one cool image. It is much harder to keep the same character recognizable while changing the world around them. That was exactly what I wanted to test with this Sonic-inspired OC.
So instead of generating three completely different characters, I kept the same core design language and placed the character into three different visual settings: a clean platform-style world, a neon-lit cyber city, and a warm desert canyon at sunset.
What surprised me most was how much the character’s identity stayed intact while the emotional tone changed dramatically from image to image.
The first image feels like the “default identity” of the character
The first version is the most neutral, and that is exactly why it matters.
The background is simple and uncluttered. The soft environment, open space, and platform-inspired elements make the character stand out immediately. My attention goes straight to the essentials: the blue body, the bright green eyes, the white gloves, and the red shoes. The pose is calm but confident, which makes this image feel like a baseline reference for the OC.
What I like most here is the clarity. Nothing is fighting for attention. The lighting is soft, the silhouette is easy to read, and the character design feels stable. If I were introducing this OC for the first time, this is the image I would use first because it defines the visual identity in the cleanest way.
This image also reminded me of something I often forget when generating AI art: not every picture needs to be visually loud. Sometimes the best first image is the one that explains the character most clearly.
The second image changes the personality without changing the character
The cyber city version was the one that felt most different emotionally.
The first thing that changed was the atmosphere. The wet ground reflections, blurred neon signs, and dark urban depth made the character feel sharper, faster, and slightly more rebellious. Even though the OC design stayed mostly the same, the environment added a whole new attitude.
Compared with the first image, this one feels more cinematic and more aggressive. The lighting is more contrast-heavy, and the background creates motion even though the character is standing still. The city lights reflect on the floor and subtly push the character into a more futuristic space. The shoes also stand out more here because the red and white colors echo the artificial glow around them.
What I found especially interesting was that the exact same face started to feel different in this setting. In the first image, the expression reads as confident and approachable. In the second image, the same confidence feels cooler, harder, and more competitive.
That is one of the things I enjoy most about AI image generation: you are not always changing the character directly. Sometimes you are changing the emotional reading of the character through the world behind them.
The third image gives the OC a stronger sense of adventure
The third image is probably the most dramatic in terms of energy.
Unlike the first image, which feels controlled, and the second, which feels urban and stylish, this one feels open, physical, and adventurous. The orange sunset, canyon depth, airborne dust, and more action-ready pose all make the character seem like they belong in a harsher world.
Get Rebecca’s stories in your inbox
Join Medium for free to get updates from this writer.
The composition helps a lot here. The stance is wider, the body language feels more prepared for movement, and the surrounding dust adds a sense of force. Even before anything happens, the scene feels like the moment right before a dash, a jump, or a battle.
The color contrast is also stronger. The blue character against the warm orange environment creates a very obvious visual tension, which makes the OC pop more than in the other two images. Instead of blending into the scene, the character actively pushes against it.
For me, this version felt the most like a story frame rather than a character showcase. It looks less like a design sheet and more like a still from a larger adventure.
What changed in each image, and what stayed consistent
This is what made the experiment satisfying: the setting changed a lot, but the identity signals stayed readable.
The quill shape remained the strongest design anchor across all three images. The eye color also helped keep the character recognizable instantly. The gloves and shoes worked like visual signatures, especially because they remained stable while the environment shifted.
At the same time, each scene changed a different layer of meaning:
- The first image defined the character.
- The second image stylized the character.
- The third image dramatized the character.
That difference matters more than people think. When I compare AI-generated images, I do not just ask which one looks “better.” I ask what each image is doing for the character.
If I want a clean OC introduction, I would choose the first one.
If I want a modern, eye-catching post thumbnail, I would choose the second.
If I want a cinematic story moment, I would choose the third.
My personal experience while generating these
What made this experiment fun was that I was not trying to chase randomness. I was trying to control variation.
A lot of AI art tests fail because the character drifts too much between generations. The face changes, the proportions shift, or the design stops feeling intentional. Here, I wanted the opposite. I wanted the same character to survive different worlds.
That changed how I looked at prompting.
Instead of rewriting everything every time, I found it more useful to lock in the character traits first and then only change the environmental layer. Once I started thinking that way, the outputs became easier to compare. I was no longer asking the model to invent a new subject. I was asking it to reinterpret the same subject.
That small mindset shift made the process feel much more like art direction and much less like gambling.
And honestly, that is the part I enjoy most now. Not just getting a pretty result, but seeing how small prompt changes can create completely different emotional outcomes while preserving the core identity of the character.
Final thoughts
This three-image test reminded me that AI image generation gets much more interesting when you stop treating each result as a one-off picture.
A single character can look clean, futuristic, or adventurous without losing who they are. That is where the real fun begins. Not just generating images, but building visual consistency across multiple worlds.
For me, that is when an OC starts to feel real.
