Blog
Blog
I write down what I'm learning as I research: automatic music transcription, signal analysis, digital lutherie.
-
DIY LabCan you separate the drone from the bagpipe in a field recording?
We fed an algorithm a recording of a real piper from Teverga, nearly 50 years old, and tried to separate the drone from the melody before analyzing it. The standard technique for this, HPSS, is what we used in this small experiment — and here's the result, honestly.
-
EssaysThe brothers who said no to Philips: the 1942 Galician tape recorder
Two Galician brothers invented a machine that recorded sound on steel wire in a family workshop, in the middle of the post-war years and with barely any electronic components available. When Philips wanted to buy the technology from them, they said no. This is the real story, with its bittersweet ending.
-
EssaysPISA 2025 Spain: AI and digital distraction, with data
I analysed the PISA 2025 microdata for Spain: digital distraction in class is associated with lower scores; AI chatbot use doesn't follow a simple pattern.
-
Music InformaticsRhythm games: music as the conductor of training
Stage Tour announces RedOctane's return. The announcement is the hook; the underlying idea —music as the conductor of a training system— predates five buttons.
-
Music InformaticsRACK-02: what's actually happening inside the techno generator
How can a program 'invent' a different techno beat every time and still make it sound good? RACK-02 does it in the browser, nothing to install. Here's how, no magic and no AI: rules, probability and a seed.
-
DIY LabGaita Hero and EMTI Fervienza: evaluating the sound
In 2011 I designed a system to check Asturian bagpipe scores. I'm still pulling the same thread: comparing a performance against a reference is an MIR problem.
-
Music InformaticsPython for music: librosa, music21, and pretty_midi
One language, three separate toolboxes: librosa looks at the audio, music21 at the score, and pretty_midi at MIDI. What each library does and when to use it.
-
Music InformaticsAsturian bagpipe organology as data
How the Asturian bagpipe can be described in measurable terms — parts, key, and frequency range — and what you need to understand before digitising it.
-
DIY LabMIDI vs MusicXML in Python: what survives and what doesn't
An experiment with music21 and pretty_midi: the same phrase exported to MIDI and MusicXML. What survives in each format, with real figures.
-
EssaysData sovereignty: from principles to architecture
How the GDPR, the AI Act and the Digital Bond Manifesto translate into technical decisions: no analytics, no cookies, with an explicit legal basis.
-
Music InformaticsWhat Music Information Retrieval (MIR) is
The field behind automatic transcription, chord recognition, and Shazam. What MIR is and how MIDI, MusicXML, and AMT fit into it.
-
EssaysAI as a tool, not as the final product
Why I see generative AI as a tool that amplifies what you already know, not a substitute for judgment. With real data, not hype.
-
DIY LabHow to compute a spectrogram with librosa
A step-by-step tutorial, with a downloadable notebook, on going from real audio to its spectrogram with librosa: STFT, decibels, and reading harmonic bands.
-
Music InformaticsWhat MusicXML is: the score as data
Moving a score between programs without it falling apart seems trivial, and it is not. What MusicXML is and how it differs from MIDI.
-
Music InformaticsWhat automatic music transcription is
Turning a recording into a score sounds trivial until you try it. What automatic music transcription (AMT) is and why it remains an open problem.
-
EssaysWhat a Jupyter notebook is and why to start with it
A Jupyter notebook blends explanation and runnable code in one document. What it is and why I think it's a great way into programming, especially for teaching.
-
Music InformaticsopenDAW: an open-source DAW that lives in the browser
What openDAW is, how it runs in the browser via the Web Audio API and TypeScript, and how its data sovereignty model connects with GDPR and free software.
-
DIY LabHow to build an audio corpus for music research
An audio corpus is not a playlist: it is a structured, labelled, reproducible collection. What it takes to make one genuinely useful for MIR research.
-
DIY LabMaking music with Arduino: from buzzer to MIDI
How to get sound out of an Arduino, from a buzzer with tone() to turning sensors into controllers. A practical walkthrough, no expensive hardware.
-
EssaysThe Fourier transform: from time to frequency
Why sound can be understood as a sum of frequencies, what the Fourier transform does with that idea, and how it explains the spectrum.
-
Music InformaticsWhat MIDI is: instructions, not audio
What MIDI is and, more importantly, what it is not. How a MIDI message describes a note without audio, and why that makes it so flexible.
Nothing here yet. Try another word — or drop me a line and I'll cover it.