field guide ยท browser audio
a web page can be a complete sound engine.
The Web Audio platform gives browsers a precise timeline, signal graph, oscillators, filters, effects, sample decoding, and offline rendering. A music generator can therefore compose and render on the listener's device.
Composition and sound are separate layers
A browser generator first creates a score or event plan: notes, sections, instruments, timing, and automation. A playback layer schedules that plan onto synthesizers, decoded samples, filters, delays, reverbs, compressors, and gain stages.
Separating the layers makes the composition testable without listening and lets one musical identity be rendered through different documented engines.
Local rendering changes the product
When composition and synthesis run locally, there is no generation queue and no need to upload the creative prompt for remote inference. Playback can begin as soon as the browser prepares the required code and source assets.
Everybeep still uses ordinary hosting for application files, optional consent-based analytics, and aggregate community signals. Its privacy page distinguishes those network requests from local musical generation.
Samples, synthesis, and procedural voices
Browser music does not have to use only oscillators. Everybeep combines procedural chip synthesis, project-original PCM sources, formant-based synthetic vocals, and nonreferential DSP systems.
Exports are rendered from explicit recipes and include technical provenance. Their availability does not imply commercial usage rights; licensing remains a separate policy decision.
Limits and tradeoffs
Browser audio must handle autoplay rules, device performance, memory limits, tab suspension, mobile interaction, and differences between browser engines. High-quality experiences prepare audio after a user gesture, cache immutable sources, bound render work, and test multiple engines.
The reward is a portable instrument: a link can contain the identity while the listener's browser performs the piece.