AI Helps Me Overcome My Impedance Mismatch — Most Of The Time

I always found Indian history and mythology far more engaging in Amar Chitra Katha comics than in school textbooks, which rarely had illustrations or photographs. Later, I got hooked on MAD magazine. It became my quirky friend, philosopher, guide, and even therapist in its own odd way. I’ve realized I learn best through visuals and light-hearted humor, not from dense textbooks packed with serious, endless text.

After being exposed to such high-quality artwork, I couldn’t help but develop the urge to draw - or more specifically, to cartoon.

But inspiration only goes so far. It takes practice and patience to actually get good at something. I often felt disappointed with what I drew and as a result I ended up drawing even less.

Impedance mismatch is an engineering term that generally refers to a disconnect or incompatibility between two components that are meant to function together. Years ago, while reading about impedance mismatch, I realized I was experiencing it firsthand. 

I could instantly visualize a cartoon in my head but translating that image onto paper never came out quite the way I imagined. 

I explored a few digital comic tools like ComicGen to keep my hobby alive.

Once in a while, my hand-drawn cartoons would turn out just the way I imagined. 

When Adobe introduced Firefly, its AI-powered image tool that lets you edit visuals using text prompts, I was thrilled! That’s when I co-created my very first AI-assisted cartoon two years ago in July 2023 without actually drawing - I reused my hand-drawn cartoon.

I often rely on multiple AI assistants to create the final output. Starting with a basic idea for which I have a visualization in mind, I provide an outline as a prompt and ask each assistant to expand on that idea by developing a scenario and setting for a single-panel cartoon. I prefer dialogues or punchlines that sound natural and have broad, universal appeal. I then choose the one I find funniest or most interesting, sometimes blending elements from the other suggestions. I enjoy the brainstorming process because it lets me discover a variety of ideas I wouldn't have thought of by myself.

I named the series "Brainstormed With Bots" reflecting my ongoing experimentation with various AI assistants like ChatGPT, Gemini, Copilot, Perplexity, and Mistral which can generate images. 

At first, they could barely generate cartoons, but they’ve all improved dramatically over time. These tools don’t just help with images, they’re also great for punching up dialogue, building scenarios, and setting backgrounds. 

Earlier versions of Microsoft Copilot openly cited OpenAI’s DALL‑E as their image generation engine, while Google’s Gemini acknowledged using Imagen under the hood. More recently, both tools have dropped the name‑brand references. Microsoft now promotes its “Copilot Designer” or built‑in image creation as a custom proprietary model without naming DALL‑E. Similarly, Google describes Gemini’s visuals as powered by a bespoke in‑house system, rather than explicitly referencing Imagen

Early on, the cartoons often had weirdly distorted body parts (which sometimes added to the humor!) 

....and constant spelling mistakes or badly rendered text. 

After the cartoon is generated, I fix minor details using simple Windows native tools like Paint 3D or Photos (when I need Generative Erase), and add text using the font MS Comic Sans, because of course. AI covers my flaws; I cover its glitches.

As the tech evolved, so did the results. AI tools now do a much better job with typography, facial expressions, and overall scene composition. 

I’ve found that ChatGPT and Copilot come closest to the look I’m after when I specify "bold lines with minimal shading" and "single panel" to keep the cartoon simple. These AI tools now read my mind with shorter prompts (although sometimes they need both prompting and prodding).  

Being a longtime MAD magazine fan, I love the classic black-and-white aesthetic, so I now stick with that style to give my cartoons a consistent feel. 

The cartoon rendering process in Copilot begins as a blurry image, with the blur gradually fading from the top downward to reveal the full illustration. The slow reveal adds a sense of anticipation and makes the process feel surprisingly exciting.

It is like seeing the moon slowly slip out from behind the clouds, first it teases you with a sliver of a shape, then the whole image reveals itself. I’m genuinely more excited than I am watching a solar eclipse! 

Most people take a photo to remember a moment. But when I hear a funny line, have an odd insight, or experience something quirky, I capture it differently...I turn it into a cartoon. My favorite AI assistants now help bring these visual ideas to life in just a few minutes - well, most of the time.