Bringing My Robots to Life: My First AI Animation Experiment
CT World | Desk Diaries | AI Animation Experiment #001
Can generative AI turn a handful of robot characters into the stars of their own animated adventure?
That's the question I decided to explore with CT-1, Wheelie and Mambo, three members of my little robot family.
I've been experimenting with robotics, artificial intelligence and virtual environments through my personal project, CT World. This time, I wanted to try something a little different: creating a cinematic 3D animation featuring my robots.
Not just a few animated expressions or simple movements, but an actual story, complete with characters, camera movements, lighting and a touch of magic.
The result is a 41-second AI-generated animation.
Is it perfect? Far from it.
But that's exactly why I'm sharing it.
The Experiment: A Mysterious Book and Three Curious Robots
The premise is simple.
One evening, CT-1 discovers a mysterious book on a desk. An unusual golden symbol glows on its cover.
Curious, CT-1 approaches the book and investigates. Wheelie soon joins the adventure, followed by Mambo.
As the three robots gather, the book begins to respond. Its mysterious glow intensifies, magical effects emerge, and something extraordinary appears to be happening.
The longer-term idea is to use this mysterious book as the gateway to CT World — an imagined universe where my robot characters can embark on adventures beyond their usual surroundings.
For now, though, this is simply an animation experiment.
Watch AI Animation Experiment #001
CT-1, Wheelie and Mambo investigate a mysterious magical book in my first cinematic 3D AI animation experiment.
How I Created the Animation
I wanted to explore whether today's generative AI tools could help an individual creator produce animated stories without a conventional animation studio.
For this experiment, I combined two AI tools.
ChatGPT — Story Development and Pre-production
I used ChatGPT to help develop the storyline, organise the sequence of events, create visual references and prepare a cinematic storyboard.
We broke the story into smaller scenes, deciding what each character should do, how the camera should move and how the magical effects should develop.
The biggest challenge was translating those plans into instructions that an AI video generator could follow consistently.
Google Gemini — AI Video Generation
I then used Gemini to generate the animated sequences.
Rather than creating one complete film in a single attempt, I experimented with generating and extending shorter sections of the animation.
This approach offered some control over the progression of the story, but it also revealed several limitations.
And that's where things became interesting.
What Worked Surprisingly Well
The first pleasant surprise was the cinematic presentation.
The AI generated atmospheric lighting, convincing materials and camera movements that gave the scenes a surprisingly polished appearance.
The warm desk environment worked particularly well with the mysterious golden glow of the book.
CT-1's cardboard appearance also translated reasonably well into animation. His square head, round glasses and simple movements helped convey curiosity and surprise.
The magical effects were another highlight.
Watching the book respond to the robots, release glowing particles and create an increasingly mysterious atmosphere was genuinely satisfying.
For a first attempt at AI filmmaking, these were encouraging results.
The Problems: When AI Has Other Ideas
Of course, things didn't always go according to plan.
1. Character Consistency Is Harder Than It Looks
Wheelie and Mambo were the biggest challenges.
My original Wheelie is a compact two-wheeled robot with exposed electronics, colourful wiring and a small digital face.
Mambo is a four-legged robotic dog with a distinctive mechanical structure and expressive screen.
Unfortunately, the AI-generated versions didn't always resemble their original designs closely enough.
Details such as body proportions, mechanical components, wheel placement and facial features could change during generation.
The resulting characters were certainly animated robots, but they weren't always recognisably my robots.
For a one-off animation, that might be acceptable. For a continuing series featuring established characters, it's a much bigger problem.
2. Getting AI to Continue a Story
One particularly frustrating challenge involved scene continuity.
When I asked the AI to generate the next section, it sometimes repeated earlier actions instead of progressing immediately to the new scene.
Characters would approach the same object again, or the animation would spend time recreating events that had already happened.
I experimented with more explicit instructions, including specifying the starting positions of characters, their actions and how the new section should connect to the previous one.
That helped clarify my intentions, but it didn't eliminate the problem completely.
3. Objects Don't Always Stay Consistent
Even the mysterious book had a mind of its own!
Its position, orientation and appearance could change between sequences.
At one point, it opened and released magical effects, only to return to a closed position later.
These inconsistencies might go unnoticed in an isolated AI-generated clip, but they become much more obvious when trying to construct a continuous story.
4. Beautiful Animation Doesn't Necessarily Mean Good Storytelling
Perhaps the most interesting lesson was that generating an attractive animated scene is not the same as directing a coherent film.
A successful animation needs more than impressive visuals.
Characters must remain recognisable. Their actions should make sense. Objects must behave consistently. Events need to progress logically, and each shot should contribute to the story.
AI can generate some impressive moments, but maintaining control over the complete sequence remains a challenge.
Why I'm Sharing an Imperfect Experiment
I could keep regenerating scenes until the characters look closer to their original designs and the storytelling becomes more consistent.
But I think there's value in sharing the process as it happens.
CT World has always been a place for me to explore robotics, AI and creative technology.
Not every experiment needs to produce a polished final product.
Sometimes the most useful discoveries come from finding out where the technology struggles, understanding why something didn't work and experimenting with better approaches.
This 41-second animation represents my first attempt at bringing these robot characters into a cinematic story.
There are mistakes, inconsistencies and plenty of room for improvement.
But there are also moments that make me curious about what might be possible with further experimentation.
What's Next for CT World?
I'd like to continue exploring the possibility of producing a small animated series featuring CT-1, Wheelie and Mambo.
My next experiments will focus on improving character accuracy, preserving visual continuity between scenes and developing a more reliable AI-assisted animation workflow.
I'm also interested in finding out how much creative control an individual filmmaker can realistically achieve with generative AI tools.
Can these tools eventually support a coherent animated series with recurring characters, consistent environments and engaging stories?
I don't have the answer yet.
But that's what makes it worth exploring.
For now, enjoy my first 41-second experiment.
Welcome to CT World — where little robots have big adventures! 🤖✨
This animation was developed with assistance from ChatGPT for storyboarding and production planning, and generated using Google Gemini. It is an experimental work in progress. The animated robot designs do not always accurately represent their original real-world references.
Frequently Asked Questions
Can AI turn real robots into animated characters?
Yes. Generative AI tools can create cinematic animations inspired by real robot designs. However, preserving exact mechanical details, character proportions and consistent appearances across scenes remains challenging.
What AI tools were used for this animation?
ChatGPT assisted with storyboarding, character references and animation prompts, while Google Gemini generated the 41-second 3D animated video.
Is CT World becoming an animated series?
That's the goal of these experiments. I'm exploring whether AI-generated animation can support a series featuring CT-1, Wheelie and Mambo while maintaining consistent characters and storytelling.
