When ChatGPT Drew the Scene: The ATM That Cried
- Nada Pleskonjic

- Apr 23
- 2 min read
Updated: Apr 28
A Breakdown in the System — and in the Method
I often use ChatGPT to reflect on ideas, edit drafts, or clarify the narrative tone — while MidJourney typically handles the visuals.
But this time, the roles reversed: ChatGPT generated the image. And not just one — a whole visual evolution.
Let me show you how we made a machine cry.
Not emotionally. Visually.
PHASE ONE: A Painterly Start
The first outputs felt like sketches — stylized, almost classical in tone.
A woman and an ATM. No real emotion. No chaos. But a good seed.


Initial renders by ChatGPT – conceptual, but too calm. The ATM isn’t breaking yet.
PHASE TWO: Enter Contrast and Realism
I asked ChatGPT to adjust the visual language:
“Make it more realistic. Boost the outfit contrast with yellow and red outfit. Bring tension.”
The result? A sharp yellow jacket. Red pants. Clearer light. But still — no breakdown. Too much fashion. Too little collapse.

PHASE THREE: Style, Absurdity, but Still Contained
We tried a more editorial tone: flowing fabric, heightened contrast, and a near-perfect composition.
Visually stunning, but still… it lacked narrative movement. The ATM looked chic. Not emotional.

PHASE FOUR: The Collapse — And the Animation Begins
Then it clicked. The final image needed to be not just stylized, but broken.
Emotionally unstable. Money flying. A woman who couldn’t stop the leak.

I gave Kling 2.0 the task: animate the scene.
Add drips. Rhythm. Tension.
Let the ATM cry.
Final Thoughts: When AI Imagines Emotion
Maybe this is about more than an absurd moment.
Maybe it’s about role reversal — when a machine tries to simulate drama, and another machine tries to draw it.
“This is not financial advice. But she tried.
And the ATM cried.”
Want more?


