KR//LAB

GenAI Audio Experimentations
Case Studies

An ongoing series of production-grade GenAI audio experiments — pipelines built and measured against director's briefs, every decision logged and every failure kept. Each lab tests one working assumption about what generative audio tools can and can't replace in a professional pipeline.

001
Designing a Prompt-Driven GenAI Audio Pipeline for Short-form Content
Can GenAI accelerate audio production while preserving creative intent and production quality? A director's brief, a vertical microdrama scene, Suno and ElevenLabs generation, an Ableton mix, and a QC pass that caught the mastering engine's own mistake.
Read the Case Study
002
Prompt-Driven vs Automatic Video-to-Audio
Four automatic video-to-audio tools, the same silent scene from 001, and the same measurements. Told nothing and told everything, then scored and listened to, side by side against a human-built reference.
Read the Case Study
003
Sonic Consistency Systems for Multi-Episode Content
A sonic identity system so episode 12 of a microdrama sounds like episode 4 — the actual scale problem content platforms face.
Upcoming
004
A Hindi Dialogue Workflow for AI Video
One word came back wrong in a Hindi AI-generated scene. The fix measured 0.25 seconds — and the files that made it possible were already on the drive.
Read the Case Study
005
What Happens To a Performance in the AI Video When You Dub It Into Multiple Languages
What survives when a performance is translated rather than regenerated. One therapy-room scene dubbed into four languages, with and without lip sync, and what the performance loses at each step.
Read the Case Study
006
Directing a Dialogue Performance in Hindi
Whispers, sighs, stretched words and misplaced emphasis. What it takes to direct an AI performance in Hindi, and why the accent lives in the spelling rather than the voice.
Read the Case Study