This is a part of a project I designed for a Mobily event.
The requirement was to take the event's general graphic branding and create 3D-simulated motion for several giant screens.
The main challenge: The client requested subtlety – not letting the AI "run wild," while maintaining slow and precise movement.
The main challenge: The client requested subtlety – not letting the AI "run wild," while maintaining slow and precise movement.
I used Sora and Nano Banana to create frames where some elements appear extruded or recessed, then moved to Veo using a start frame and an end frame.
I used GPT to write dozens of prompt variations to fine-tune the motion dynamics.
I used GPT to write dozens of prompt variations to fine-tune the motion dynamics.
Needless to say, it took several hundred iterations... but I have reached my destination..
This was the original design
Some of the start and end frames I created with Sora
