This AI Cleans Up Your Audio Better Than You Can

You recorded a vocal take in a room with street noise. Or your podcast has HVAC hum you didn't notice until editing. Or a film dialogue track is buried under ambient sound. You could spend hours cleaning it up manually. Or you could run it through Hance.

Peder Jorgensen built Hance.ai from a studio in Oslo -- a machine learning audio processor that removes noise from dialogue in real time. It runs on hardware as light as a Raspberry Pi. If you've watched a recent film, there's a good chance Hance technology touched the audio.

Why This Matters for Your Studio

Hance isn't another plugin competing in the crowded music production market. It targets the professional audio world -- film post-production, broadcast, sound design. The sweet spot where people actually have budgets and deadlines.

The technical edge: Peder and his co-founder wrote their ML models from scratch, specifically for audio. Not adapted image recognition models like most competitors. Their musical background matters -- Peder has seen training datasets recorded at "32-bit 16 kilohertz, which is just super weird if you understand audio."

Their live stem separation can pull vocals from a mixed track in real time. That started as office karaoke parties and is becoming an actual product.

What You Can Learn From Peder's Approach

Build for yourself first. Peder's other company, Soundly, came from his own workflow problem as a sound designer. In 2012, the industry was still buying CDs of sound effects. He built a cloud library for himself, then watched it spread through Oslo's sound design community. Entirely bootstrapped, entirely profitable.

Pick the right niche. The music plugin market is enormous but brutally competitive because it's passion-driven. Film and sound design is smaller but more professional -- people have real jobs, real budgets, and real problems to solve. That stability let Peder bootstrap without investor pressure.

Domain expertise is your moat. In AI, understanding the data matters more than understanding the algorithms. A music producer who learned to code will build better audio AI than a software engineer who listened to a few albums.

If you run a studio, the same principle applies. Your ears, your room, and your experience are things no AI can replicate. Tools like Hance handle the grunt work. You handle the taste.

Source: Podcast, Episode 30 -- Peder Jorgensen (Hance.ai / Soundly)