The Kalank cover was supposed to be just a vocal take. I recorded it, listened back, and kept seeing things — a wanderer on a road that doesn't end, a spirit made of golden light, a bridge that shouldn't be able to exist. So instead of shipping the audio, I started building what I saw.
That became my first fully AI-animated music video, and a process I now use for every release.
Music first, always
The visuals never lead. I listen to the finished vocal on repeat and write down what each section feels like — not what it should look like. "Ranjha" gave me longing, devotion, a search. Only after the emotional map is done do I translate feelings into imagery: the longing became the wanderer, the devotion became the golden spirit, the search became the bridge.
If you start with "cool shots," you get a showreel. If you start with the song, you get a film.
Scene by scene, frame by frame
Every scene is a conversation between what I hear in the music and what I can make appear on screen. My working loop:
- Storyboard in words. One line per musical phrase. The chorus gets the widest shots; the quiet verses get faces and hands.
- Generate stills until the world is consistent. Character, palette, light. Consistency is the hard part of AI filmmaking — I lock a visual grammar (golden light, deep blues, dusk skies) before animating anything.
- Animate the keyframes. Motion follows the vocal line. When the melody rises, the camera lifts. When I hold a note, the frame holds too.
- Cut to the breath. I edit on the vocal phrasing, not the drum grid. It's a singer's cut, not a DJ's.
Why this matters to me
Independent musicians have always had the songs but never the budgets. A cinematic video used to cost more than an entire EP. Now the constraint isn't money — it's taste, patience, and how clearly you can see what the song already contains.
"Every song's a frame" isn't a tagline for me. It's literally the workflow.