Vyanjan: Code & Data Kit (Music Video in Python) with Dhananjay Aggarwal

Ever watched a music video and wondered how you'd build one entirely in code? This kit hands you the full engine behind one.

I made "Vyanjan", a 3.5-minute music video for KR$NA's track where all 12,594 frames are drawn in Python. No After Effects, no stock footage. The blog post shows you the method. With this kit, you get the real thing: every line of code, every number and every plan behind it.

WHAT YOU GET

📦 The full renderer: 11,220 lines of Python across 41 modules. All 33 letter scenes plus the intro and finale, camera paths, springs, line-art 3D (train, conch, heart, solar system), a veena string simulation, Kuramoto fireflies and correct Hindi typesetting. You can render one frame, any clip, or the whole film.

🎵 The audio data: 60 fps kick, snare, hat and vocal tracks from separated stems, a 337-beat grid, 707 word events and all 33 letter timings. As NumPy and plain CSV, so you can also use it in After Effects, Blender or TouchDesigner.

📖 The docs: about 16,800 words explaining every scene's science, numbers, camera and transitions, plus the quality gates every shot had to pass.

📝 Concept sheets: how each chapter was planned, with beat tables, bridges between letters and camera plans. You see the thinking, not just the code.

🔬 The analysis pipeline: the stem separation and 60 fps feature scripts, so you can turn songs you have the rights to into data the same way.

WHAT YOU CAN DO WITH IT

• Re-render Vyanjan frame for frame on your own machine

• Swap any scene for your own idea

• Take the cameras, springs, bloom, data panels and Hindi text rendering into your own projects

• Learn audio-reactive visuals from working code instead of tutorials

• Skip the long trial-and-error behind the timing, cameras and transitions

BLOG VS KIT

The blog gives you about 60 lines of snippets and 7 stills. The kit gives you the complete 11,220-line engine, every scene, all the data and the docs.

WHAT YOU NEED

Python 3, ffmpeg and libraqm. Tested on macOS (Apple silicon), with Ubuntu/Debian setup in the README. No GPU needed. The full film renders in about 26 minutes on an M1.

This is an independent work, not affiliated with the artist or label.

Personal licence for one person. Stuck on setup? Email me at [email protected] and I'll help you out.