A Decade in the Making
drmt isn't a startup pivot or a weekend hackathon, it's the culmination of ten years thinking about AI and audio, built by someone who wrote the book on it.
Dr Michael Terrell
Founder, drmt
The Story
This has been a passion project for over a decade. I wrote my PhD thesis on AI music production, working out how machine learning can understand and enhance audio. It wasn't a hobby or a side project, but was the focus of years of academic research.
I was an early researcher at LANDR when the category of AI mastering was being invented. I saw first-hand what worked and what didn't, and have honed my perspective on how these systems should work: they should be fast, they should be transparent, and they shouldn't get in the way of existing workflows.
Since then I've led data teams at Channel 4, The BBC, and TalkTalk, learning how to build real-world systems that work at enterprise scale, not just in research papers.
Now I've finally built what I've been thinking about for years. A mastering tool that's fast, transparent, and built on genuine research.
Why We're Different
Speed
Designed from the ground up for speed with novel algorithm approaches, rather than ML bolted onto a traditional audio stack.
Transparent
Every master returns the DSP parameters that produced it, helping users to learn and understand what's happened, and providing a fully auditable service.
Seamless Integration
REST APIs and batch endpoints, along with the speed, mean that mastering happens inside your platform and without interrupting your customer's flow.
How does it work?
We've taught our models what great mastering actually sounds like, by training them on tens of thousands of professional recordings. Upload a track and they go to work: EQ, compression, stereo balancing, limiting, all dialled in to match the characteristics of the pros. The breakdown below is from a real master. You'll see yours laid out the same way, every time.
EQ
Adjusts the tonal balance across the frequency spectrum — a well-applied EQ ensures your mix translates across all playback systems, from earbuds to studio monitors.
EQ
Compression
Subtle mastering compression glues the mix together, adds punch and energy, and ensures consistent playback levels.
Compressor
| Threshold | -27.7 dB |
| Ratio | 2.5:1 |
| Attack | 79 ms |
| Release | 143 ms |
| Avg GR | -4.1 dB |
| Peak GR | -8.7 dB |
Stereo Image
Shapes the spatial image of your mix across the frequency range, often widening the mids and highs for an immersive sound, while centering bass frequencies to ensure your track sounds solid on mono systems like club PAs and phone speakers.
Stereo Image
Limiting & Loudness
Applies gain to bring the track up to the target loudness level, ensuring your track holds its own alongside commercial releases while preserving dynamics and preventing clipping.
Limiter & Loudness
| Gain | +1.5 dB |
| Ceiling | -0.1 dB |
| Peak | -0.1 dB |
| Limited | 2.0% |
| Peak GR | -1.5 dB |
| Input | -19.8 LUFS |
| Output | -14.0 LUFS |
| Boost | +5.8 dB |
Every master includes this complete breakdown so you can learn from what the AI decided and understand why your track sounds the way it does.
Credentials
Academic
-
PhD Music Technology
AI Music Production focus
-
PhD Aerospace Engineering
Computational methods
-
Early researcher at LANDR
When AI mastering was invented
Industry
-
Head of Data, BBC
Enterprise-scale AI deployment
-
Head of Data Science, Channel 4
ML products at scale
-
10+ years in AI & ML
Research to production