Blog

From the Mozrat team

Production workflows, AI audio, and the business of music for agencies and studios.

Digital AI stems merging with analog live recording waves into a unified mix
Production Fundamentals

Mixing AI-Generated Stems with Live Session Recordings: A Practical Guide

AI stems and live recordings come from different signal chains. Getting them to sit together cleanly requires understanding where the differences actually are — reverb, dynamics, spectral space, and groove.

Priya Nair
Legal frameworks meeting audio waveforms for AI music copyright
Licensing

AI Music Copyright in the UK and EU: What Changed in Early 2026

The UK IPO clarified its computer-generated works position. The EU AI Act's transparency requirements produced their first enforcement actions. Here's what it means for production teams using generative audio commercially.

Sundar Arvind
Layered audio stems branching across game states in an RPG
Game Audio

Adaptive Audio for a 40-Hour RPG: What We Learned Delivering Stems at Scale

1,400 stem files across 80 zones, 4 combat tiers, and 12 ambient states. Here's what broke, what we fixed, and what we'd do differently on the next project of this scope.

Sundar Arvind
Audio waveforms racing through a latency measurement pipeline
Technical

Latency Benchmarks for Real-Time Stem Generation: Where We Are Now

We instrumented every stage of our generation pipeline and benchmarked across four output duration tiers. P50 numbers, where the time actually goes, and what real-time threshold means in practice for production workflows.

Priya Nair
What production professionals want from AI music in 2026
Industry

What Production Professionals Actually Want from AI Music in 2026

Not a library. Not a DJ. Professionals want: stems, sync clearance, revision granularity, and API access. Here's what we heard from agencies, game studios and editorial teams over the past year.

Sundar Arvind
Podcast background music workflow using stem separation
Content Creators

Podcast Background Music: How Stem Separation Solves the Ducking Problem

Ducking background music under a voice track works perfectly when you have the isolated rhythm and melody stems. With a mixed master, you're fighting the full frequency stack. Here's the difference in practice.

Sundar Arvind
Technical deep dive into harmony stem isolation in AI music generation
Technical

How We Isolate the Harmony Stem: A Technical Walk-Through

Harmony stem isolation is the hard part. Melody and rhythm separate cleanly from frequency analysis. Harmony — chords, pads, tension — bleeds into both. Here's how our model approaches the problem.

Priya Nair
Music revision cycle in advertising production — time and budget impact
Advertising

The Music Revision Cycle: Why It Kills Ad Campaign Schedules and What to Do About It

The first draft is never what the client wanted. The third draft is what they agreed to. That revision gap eats 4–6 days from a campaign schedule — here's how production teams are starting to close it.

Sundar Arvind
Game audio budget breakdown for indie and AA studios
Game Audio

Game Audio Budget Breakdown: Where Indie Studios Actually Spend Their Audio Money

Based on conversations with 12 indie and AA game studios: what goes in the audio budget, what gets cut first, and how adaptive music ambitions collapse against composer availability.

Sundar Arvind
WAV versus MP3 audio format quality in post-production
Production Fundamentals

WAV vs MP3 in Post-Production: Why Lossy Audio Is a Problem Your Editor Will Find

MP3 artefacts survive the first export. They do not survive three generations of conform and recompress. Here's why 24-bit WAV stems are the non-negotiable baseline for professional delivery.

Priya Nair
Comparing composer cost vs AI music generation cost
Business of Music

Composer vs AI Music Generation: The Real Cost Comparison Nobody Publishes

A mid-tier composer revision cycle costs £1,800–£4,500 for a 60-second spot. Here's what the actual cost structure looks like, and where AI generation changes the economics.

Sundar Arvind
Writing an effective brief for AI music generation
Workflow

How to Write a Music Brief That AI Can Actually Use

Tempo, mood, reference track, instrumentation cues. The difference between a brief that produces generic output and one that produces something usable comes down to three specific inputs.

Priya Nair
Content ID claims and stock music — the licensing problem for creators
Content Creators

The Content ID Problem: Why Stock Music Creates Claims Even When You Paid for It

You licensed the track. YouTube still claimed it. Here's why Content ID hits licensed stock music and what original generation actually protects you from.

Sundar Arvind
Adaptive music in game middleware using stem separation
Game Audio

Adaptive Music in Wwise and FMOD: Why Stem Separation Is the Starting Point

Most adaptive music middleware assumes you have separated layers — melody, harmony, rhythm. But most composers deliver mixed masters. Here's the workflow gap and how to close it.

Priya Nair
Sync licensing for advertising producers — contract and creative
Licensing

Sync Licensing 101 for Advertising Producers: What You're Actually Paying For

Master use, sync rights, blanket licenses — the language of music clearance is designed to confuse buyers. This is what actually matters when you're clearing a track for a broadcast campaign.

Sundar Arvind
What are audio stems and why production teams need separated tracks
Production Fundamentals

What Are Stems? Why Production Teams Need Separated Tracks, Not Mixed Masters

A stem is an isolated audio layer — melody alone, harmony alone, rhythm alone. Here's why the difference between a stem set and a mixed master changes everything about your post-production workflow.

Sundar Arvind