Work / Recent
AI-assisted visual direction · music video
Mesaj / Message
An early AI-assisted music-video experiment: generated character and scene imagery developed into moving sequences, then selected and edited into a coherent visual world for Duygu Soylu.
Intent
Mesaj was my first serious experiment with AI-generated imagery as raw visual material for a music video.
The aim was not to demonstrate the tool. It was to build a coherent visual atmosphere for the song: warm light, summer, softness and a slightly unreal memory-like quality.
Role
Concept development, AI-assisted character and scene generation, visual direction, image-to-video experimentation, selection, sequencing and final assembly were handled by me. The song is by Duygu Soylu.
Process
The workflow moved from still visual development into generated motion and then into conventional editorial work. I first searched for a character and scene language that belonged to the song, used those directions to create moving fragments, then treated the generated footage as source material rather than finished shots.
Generative output is abundant; direction is scarce. Most of the work was therefore in rejecting material that broke the visual language, finding continuity between fragments and shaping the final timing around the music.
Reflection
The useful part of generative tools is not that they remove decisions. They create more material and therefore make editorial decisions more important. Mesaj became an early study in how I wanted to work with machine-generated imagery without letting the tool become the subject.