Beat-Sync Technology

Beat sync is the technology that aligns visual events to the rhythmic structure of music. When a particle burst fires on exactly the first beat of a bar, when a camera cut happens precisely on a snare hit, or when a bloom effect peaks at the exact moment of a bass drop, that is beat sync in action. Without beat sync, visuals react to audio with variable latency or arbitrary timing, creating a disconnect that viewers perceive immediately. Humans are extraordinarily sensitive to audio-visual synchronization; even a 30-millisecond offset between a beat and its corresponding visual event is noticeable. Beat sync technology ensures that this alignment is not approximate but exact, down to the individual video frame. This precision is what separates professional-quality music visualizations from amateur ones.

A beat grid is a timeline of precisely placed beat positions that spans the entire duration of a song. It is constructed during the audio analysis phase by first detecting the song's BPM (beats per minute) and then determining the phase (the exact timestamp of the first beat). From these two values, every subsequent beat position can be calculated. The grid also encodes higher-level musical structure: beats group into bars (typically 4 beats per bar in 4/4 time), bars group into phrases (commonly 4 or 8 bars), and phrases form sections (verse, chorus, bridge). This hierarchical structure allows visual events to be placed not just on beats but on musically meaningful positions: a transition on the first beat of a new 8-bar phrase, an effect activation on beat 3 of every bar, or a camera change every 16 beats.

Quantization is the process of snapping a time value to the nearest position on the beat grid. If an onset is detected at 2.487 seconds and the nearest beat is at 2.500 seconds, quantization shifts the event to 2.500 seconds, ensuring it lands squarely on the beat. Quantization resolution determines the finest grid division used: 1/4 note quantization snaps to beats, 1/8 note to half-beats, 1/16 note to quarter-beats, and so on. Finer quantization preserves more rhythmic detail but can also preserve timing inaccuracies. Zeltor Labs uses beat grid snapping (via `BeatGrid.snap()`) as a core mechanism for placing visual events. This quantization ensures that visual timing is always musically correct, even if the underlying onset detection has minor timing variations. The result is visuals that feel locked to the groove of the song.

Deterministic timing means that the same audio input always produces the same visual output, frame for frame. This is achieved by using the audio timeline as the sole time reference for rendering. Instead of using the system clock (`Date.now()`) or wall-clock time, the renderer queries the audio playback position (`audioContext.currentTime`) to determine what visual state to display. For video export, the renderer advances the audio position by exactly one frame duration per step, independent of real-world time. This eliminates all sources of non-determinism: frame rate variations, system load, GC pauses, and timing jitter. Zeltor Labs treats determinism as a core architectural principle. Combined with seeded random number generation (no `Math.random()`), this guarantees that a preview and its corresponding export are pixel-identical.

Implementing beat sync requires careful coordination between the audio analysis, event scheduling, and rendering systems. The audio analyzer produces a beat grid data structure containing an array of beat timestamps, BPM value, time signature, and section boundaries. The event scheduler consumes the beat grid and resolves visual events to specific frames: "trigger effect X at beat 32" becomes "trigger at frame 1920 (at 60 FPS)." The renderer uses the audio clock to look up the current beat position and fractional beat progress (e.g., "we are 0.73 of the way through beat 47"). This fractional value enables smooth interpolation: a shape can scale up from beat start to beat midpoint, then scale back down, creating a bounce that is perfectly aligned to the rhythm. All of this logic runs deterministically, ensuring reproducible output across preview and export.

Beat sync is straightforward for four-on-the-floor electronic music at a steady tempo, but edge cases abound. Tempo changes within a song require the beat grid to have variable spacing. Time signature changes (switching from 4/4 to 3/4) require the bar structure to adapt. Swing and shuffle rhythms place off-beat events at non-uniform positions. Songs with rubato (flexible timing) may not have a meaningful beat grid at all. Live recordings often have gradual tempo drift that makes a fixed BPM grid progressively lose sync. Advanced beat sync systems handle these cases with variable-tempo beat grids, adaptive grid algorithms, and confidence annotations that flag regions where the grid may be unreliable. For most contemporary music, however, the steady-tempo beat grid model works extremely well.