Oğuzhan Koç Yanımda Olsan

Workflow

1. 01

   From concept to final master

   Visual concept and character development → scene images → moving-shot generation → shot selection and editing to the music → colour and transitions → final checks and Full HD master export.

Tools and models

What was human, what was generated

Leopati developed the visual concept and scene direction, selected shots, edited them to the music, graded the colour and carried out the final checks. Visual scenes and moving shots were created with generative AI models. The existing song recording supplies the music; the AI contribution is to the visual production.

Prompt approach

Character references, wardrobe, location, lighting and camera movement were defined together for each scene. Facial appearance, atmosphere and motion continuity were checked when selecting shots. The three short texts below are example prompts prepared to describe scenes in the music video. The archive contains 47 images and 63 unique video outputs; one duplicate copy, the final master and source music files were excluded from the production counts.

Use the same reference couple. A woman in a dusty rose cardigan and a man in a cream cardigan read together on a sofa. Warm window light, natural skin texture, intimate cinematic framing, 35mm lens, gentle camera push-in. Preserve faces and wardrobe; subtle movement, stable hands, 16:9, no text.

Keep the reference characters and clothing consistent. At a sunset beach, the woman raises a vintage camera to photograph the man. Amber sky, gentle waves, soft lens flare, shallow depth of field, natural expressions. Slow lateral camera movement; preserve facial identity, correct fingers and camera shape, cinematic 16:9, no text.

The same male character sits alone beside a wet tram-lined city street, wearing the cream cardigan and dark trousers. Muted stone and charcoal tones, dusk reflections, restrained melancholy, realistic skin, 50mm lens. Slowly move closer while the background stays coherent; preserve face and wardrobe, no extra people in foreground, 16:9.

Excerpts, trimmed, in the language they were written in.

Problems and solutions

1. Problem Maintaining consistent facial appearance, wardrobe, lighting and a shared atmosphere for the same characters across different locations and AI-generated shots.

   Solution We carried the characters through repeated clothing and scene references, checked the outputs for facial, motion and lighting continuity, then selected suitable shots and completed the sequence with editing, transitions and colour grading to the music.