Drops & Colors
AI-native material study
· liquid physics
· controlled randomness
This series started as a revisit of my early 2024 experiments in Adobe Firefly.
I wanted to test how far those old ideas could go today — not as pure generation, but as a systematic pipeline.
I wanted to test how far those old ideas could go today — not as pure generation, but as a systematic pipeline.
I rebuilt the visuals using a hybrid workflow across Krea, Kling AI, Grok, and OpenAI ChatGPT for prompt design and structure.
The goal wasn’t “pretty liquids”.
It was research:
It was research:
— how drops behave like objects
— how paint becomes architecture
— how foam and bubbles create depth and scale
— how to reach photographic realism while staying graphic and abstract
— how paint becomes architecture
— how foam and bubbles create depth and scale
— how to reach photographic realism while staying graphic and abstract
I treated each frame like a material study:
glass, gel, resin, soap, pigment, air.
glass, gel, resin, soap, pigment, air.
Real light logic.
Macro optics.
Physical thickness.
Soft shadows.
No “AI look”.
Macro optics.
Physical thickness.
Soft shadows.
No “AI look”.
The story evolved around a simple metaphor:
a drop as a character — colliding, merging, dissolving, leaving traces like paint on a surface.
a drop as a character — colliding, merging, dissolving, leaving traces like paint on a surface.
Technically, this project is about:
prompt-to-image transfer between tools
maintaining consistent material language
controlled randomness
bridging product photography aesthetics with abstract composition
building an AI-native pipeline instead of single-tool generation
Visually, it sits somewhere between:
cosmetics lab × fluid physics × graphic poster design.
cosmetics lab × fluid physics × graphic poster design.
Here are images from AdobeFirefly that I used.
Thank you for watching!