Case study: AI MUSIC VIDEO
This project was created by Jasmine Fernando, Müberra Daldal, Yerlin Marissen Blanco as their final project for the course Motion Design Lab
Sense
Framing Context
The project began with a collaboration with artist Kainat, who provided the song ‘Down Bad’. The story focuses on a young woman obsessing over a boy who has left her. The music video aims to visualize this emotional obsession through a symbolic journey across multiple AI-generated dream worlds. Initially, a ‘red thread’ was envisioned to represent the connection between them, which evolved into a ‘black shadow’ symbolizing lingering obsession.
Framing Goals & Intent
The team defined a clear artistic and narrative goal: create a cohesive AI-driven music video that reflects Kainat’s emotional journey. Scenes would be generated with different AI tools and connected through a consistent storytelling device. The challenge was maintaining visual and narrative coherence while working with multiple generative AI tools.
Sample
Exploring Ideas
Inspired by the idea of traveling through different eras and realities, the team created a concept in which Kainat is guided by a red thread (later black shadow) through fantasy-like AI-generated worlds:
– Tokyo Night (Müberra)
– Fairytale/Flower Field (Jasmine)
– Anime World (Yerlin)
The opening and closing scenes were to be filmed in Kainat’s real bedroom, transitioning to her dream world. Moodboards were created to define visual language per world.
Prototype
Various AI tools were explored:
– Midjourney (concept art and moodboards)
– Pika Labs (image-to-video tests)
– LensGo (training a model based on photos of Kainat for greenscreen composition)
– RunwayML (text-to-video animation with strong visual output)
– Perchance.org (initial character tests)
Each team member focused on their own world, experimenting with prompt engineering and consistency. Attempts were made to integrate the symbolic thread into all scenes, but this proved technically difficult.
Shape
Curating and refining
After testing various tools, RunwayML was selected for the final visuals due to its flexibility and visual quality. The team generated most scenes using the ‘text to video’ feature. Pika Labs and LensGo were tested but presented limitations in speed or style. A consultation with AI expert Vince Buyssens helped improve prompt quality and consistency.
Stage
Integrating Result
The final music video combines real footage of Kainat with AI-generated dream sequences. A black shadow now leads her through various emotional landscapes, symbolizing her obsessive thoughts. The workflow integrated AI tools for visual generation and manual editing for scene transitions.
Deliverables included:
– Concept development and visual narrative
– Scene-by-scene shotlist and moodboards
– Storyboard and AI-generated video assets
– Final edit prepared for exam presentation