When the Sky Writes the Score

9 November 2026 · ~9 min read · Sonification / Generative music

A mint-and-coral atmospheric landscape where wind and water become sound-wave contours.
LUCKY — live weather translated into sound

Most music begins with a person deciding what should happen next. LUCKY begins with a forecast. Wind, temperature, humidity, cloud and rain arrive as numbers; the browser turns them into tempo, harmony, texture and rhythm. Choose a place, press play, and the sky writes a score that exists only for that moment.

Weather is already a score

A weather report is a compact description of movement. Air accelerates, pressure changes, water gathers and falls, light disappears behind cloud. The usual interface turns all of that into icons: a sun, three drops, an arrow. Useful, but silent.

LUCKY asks a different question: what if those measurements were not displayed but performed? Not as a gimmick where rain triggers a rain sample, but as a small generative system in which weather changes the musical decisions themselves.

The distinction matters. LUCKY does not play a recording assigned to Warsaw or loop a track called “storm”. It builds sound in real time with the Web Audio API. The current conditions select and continuously shape the voices, filters, envelopes, delays and patterns. Return later and the same city has changed because the input has changed.

Five numbers become musical behaviour

The data comes from Open-Meteo. For the selected coordinates LUCKY reads temperature, wind speed and gusts, relative humidity, cloud cover, precipitation and the probability of precipitation. Raw values are not notes, so each has to be mapped onto something the ear can follow.

WeatherWhat it tends to change
Wind and gustsTempo, pulse rate, noise movement and rhythmic intensity
TemperaturePitch range, brightness, scale choice and mood
HumidityReverb depth, resonance and the sense of acoustic space
Cloud coverDensity, darkness, pad level and harmonic texture
RainPercussion, extra layers, pattern activity and occasional thunder

These are tendencies, not one global formula. A strong wind over the ocean should not produce the same result as strong wind in Berlin or on Kamchatka. The value is the same; the musical meaning is local.

Why one universal patch was not enough

The first obvious design is a single synthesiser: wind always controls BPM, rain always controls percussion, temperature always transposes the scale. It is easy to explain and quickly becomes boring. Every place is reduced to the same instrument with slightly different knob positions.

LUCKY instead has twenty location profiles. A rainforest can turn precipitation into pattern density. Amsterdam can let wind push an acid line and humidity widen its space. Iceland can use gusts as geothermal energy. Warsaw can borrow the pulse of a metro system; Antarctica can leave far more room around each event. The weather remains the author, but each place gives it a different ensemble.

This is not an attempt to claim that a city has one true sound. It is closer to film scoring: a deliberate musical lens placed over real data. The profile provides memory and character; the weather provides the performance.

Sonification, not simulation

There is no scientific law saying that seventy percent humidity equals a longer reverb tail. The mapping is expressive. LUCKY belongs to sonification because data controls audible parameters, but the goal is not to let a researcher read an exact temperature from a chord. The goal is to make change perceptible.

That required restraint. Direct mappings often become unpleasant at the edges. A storm should feel more active, but it should not drive the tempo beyond recognition. A heatwave may brighten a patch without pushing every oscillator into the top octave. Values are normalised, limited and sometimes combined, so the output stays musical while the input remains legible.

The most useful test was not “does this sound like rain?” It was “can I hear that the weather changed without looking at the panel?” If cloud cover rises and the texture grows denser, or a gust wakes a dormant layer, the mapping is doing its job.

The browser is the instrument

Everything runs in the page. The Web Audio API supplies oscillators, filtered noise, gain envelopes, delays and reverberant spaces. Small schedulers create rhythm and melody. An analyser draws the moving visual response. There is no audio file to download before playback and no track being streamed from our server.

The browser also imposes a useful rule: sound may begin only after a deliberate user action. That is why LUCKY asks you to choose a location and press play. The pause between seeing the map and hearing it is not just a permission dialog. It makes listening a decision.

Keeping the system client-side also means the weather request can go directly to the public data source. LUCKY needs no account, stores no musical profile about you, and can stop as completely as closing the tab.

Randomness needs boundaries

Generative music often uses randomness as a synonym for variety. Unrestricted, it is usually a synonym for drift. Notes wander out of the harmony, too many events coincide, or a quiet passage becomes permanently busy.

LUCKY uses probability inside a structure. A profile defines a palette and behaviour; weather changes the odds, rates and ranges inside it. Rain may make an event more likely, but it does not replace the whole composition with dice. The system stays recognisable long enough for a change in conditions to mean something.

This also explains the name. Luck is present, but it is never alone. Chance meets a place, a set of constraints and a live atmosphere. The result is neither a fixed song nor noise. It is a performance with rules.

A map of impossible listening positions

Twenty locations turn the project into a peculiar radio. In a minute you can move from a European city to the Amazon, a desert, Antarctica or a volcanic landscape. You are not hearing microphones in those places. You are hearing the same categories of atmospheric measurement interpreted through different musical systems.

That abstraction is the interesting part. A webcam shows what somewhere looks like. Field recording captures what happened near a microphone. LUCKY performs what the air is doing as data. It is less literal, but it makes comparison possible: how does the current wind in one place alter rhythm beside the wind somewhere else?

The honest limitations

Weather data is measured and modelled at a geographic scale. It is not a sensor outside your exact window. Updates arrive in intervals, not continuously, and a forecast service can be unavailable. LUCKY therefore represents the latest data it can retrieve, not an infallible live feed.

The location profiles are authored interpretations. They carry our taste, shortcuts and cultural associations. That subjectivity is not hidden behind a claim of objectivity; it is the compositional layer of the work.

Finally, a browser synthesiser is not a finished album. It can surprise, settle into a convincing atmosphere and occasionally produce a moment worth recording. It can also repeat itself. The project is an instrument and an experiment, not an endless supply of completed songs.

Listen to where you are

Pick the city nearest to you first. Listen without opening the weather app. Then look at the values and follow the arrows from measurement to sound. Try somewhere with very different conditions. The interface shows the mapping because the mechanism is part of the piece, not a secret behind it.

Tomorrow the same button will play something else. That is the whole idea: the work is never stored in one final version. It waits for the sky to finish the composition.

Sources

  1. Open-Meteo supplies the forecast and current-weather variables used by the instrument.
  2. Web Audio API documentation for the browser audio graph used to synthesise and analyse the sound.
Open LUCKY →