Audio Processing Science: Reverb Algorithms, Time-Stretching, and Browser Audio Engineering

1. How Browser Audio Processing Works

Browser-based audio processing is powered by the Web Audio API — a high-level JavaScript API developed by the W3C Audio Working Group. The API provides a complete audio processing graph architecture that achieves studio-quality effects entirely within the browser.

The processing graph consists of interconnected AudioNode objects: AudioBufferSourceNode (loads and decodes audio), ConvolverNode (applies convolution reverb), GainNode (controls wet/dry mix), BiquadFilterNode (EQ and bass boost), and AudioDestinationNode (output). Internal precision is 32-bit floating point, providing approximately 1500 dB of dynamic range.

2. Convolution Reverb: How It Works

The ConvolverNode performs real-time convolution using the overlap-add method, which divides audio into frames, convolves each frame with an impulse response, and reconstructs the output. For algorithmic reverb, the processing chain includes cascaded comb filters and all-pass filters based on the Schroeder reverb model (1962), with modern enhancements for diffusion density.

Given the same impulse response and dry signal, convolution reverb produces bit-for-bit identical output in a browser and a DAW. There is no measurable quality difference for reverb processing tasks.

3. Time-Stretching and Pitch Shifting Algorithms

When you slow or speed up audio, the browser must resample the waveform. SlowedReverbMaker.net uses coupled resampling (varispeed) — slowing the playback rate naturally lowers pitch, just like DJ Screw's turntable technique. The pitch shift control uses detune parameter on the AudioBufferSourceNode, which shifts pitch in cents (100 cents = 1 semitone).

For advanced time-stretching that separates speed and pitch, WSOLA (Waveform Similarity Overlap-Add) algorithms analyze the waveform, identify similar segments, and overlap them to change duration without shifting pitch. This is what professional DAWs use for independent tempo and pitch control.

• Coupled resampling (varispeed): speed and pitch change together — identical to vinyl/turntable physics.
• WSOLA: overlaps similar waveform segments to change duration without pitch shift.
• Pitch shift (detune): shifts all frequencies by a fixed ratio — +100 cents = +1 semitone.
• Low-shelf EQ (bass boost): amplifies frequencies below 150Hz with a gentle slope.

4. Quality Benchmark: Browser vs DAW

Signal-to-Noise Ratio (SNR): Web Audio API's 32-bit float processing provides ~1500 dB theoretical SNR. At 16-bit WAV export (96 dB SNR), browser and DAW processing are indistinguishable.

Total Harmonic Distortion (THD): effectively 0% — all operations are linear (filters, delays, gains). THD for browser reverb processing is identical to DAW-based processing.

Frequency response: convolution with identical IR produces mathematically identical output. Algorithmic reverbs may vary between implementations, but the difference is algorithm design, not platform capability.

5. Psychoacoustics: Why Slowed Reverb Feels Emotional

Three changes happen simultaneously: tempo, pitch, and space. Slower speed (60-80 BPM) activates the brain's default mode network, linked to introspection and nostalgia. Lower pitch (2-4 semitones) increases nucleus accumbens activity, the brain's reward center.

Reverb adds spatial depth that simulates acoustic space. A longer decay time (2-5 seconds) triggers associations with large halls and cathedrals, creating a sense of awe. Together, these effects produce the signature emotional response of slowed reverb music.

• 60-80 BPM: activates default mode network (introspection, daydreaming, nostalgia).
• 2-4 semitone pitch drop: increases nucleus accumbens activity (pleasure/reward).
• 40-80Hz sub-bass: triggers tactile entrainment — heart rate and breathing sync.
• Reverb decay 2-5s: simulates large acoustic spaces, triggers awe response.

6. Gain Staging for Clean Browser Processing

Gain staging is critical for distortion-free audio. Bass boost adds +3 to +6 dB gain. Reverb adds energy that increases RMS. If cumulative gain exceeds 0dBFS, clipping occurs.

Best practice: normalize source peaks to -6dB before uploading to a browser tool. This provides 6dB of headroom for processing. Export at moderate level and use a normalizer only if the platform requires louder files.

• Normalize source to -6dB peaks before browser processing.
• Boost conservatively: +3 to +6 dB bass, 35-45% reverb wet.
• Preview the loudest section before exporting.
• Export WAV for archival (lossless), MP3 320kbps for sharing.
• Desktop processing time: 4-5 seconds for 3-minute track. Mobile: 8-12 seconds.