Audio Trimmer

How It Works

The pipeline is decode → select → resample to 16-bit integers. Your file is read into memory and handed to AudioContext.decodeAudioData, which turns any codec the browser understands into a raw PCM AudioBuffer. The two handles define a selection in seconds, snapped to 10-millisecond steps, and export copies exactly that slice of PCM, applies the fade ramps, converts each float sample to a signed 16-bit integer and writes it into a hand-built WAV header.

Why the waveform shows peaks, not samples

One canvas column stands for thousands of samples, so the renderer stores the minimum and maximum of every column instead of individual values. A min/max pair keeps quiet passages visible and loud transients honest — downsampling to plain averages would make percussive edits look flat in exactly the places you need to see.

Why the output is WAV

WAV is a lossless container: after decoding, writing the same PCM into a 16-bit WAV adds no codec generation loss. MP3 would require shipping an encoder into the page and re-introducing lossy compression on top of the decoder's approximations, so it is deliberately not included — trim here, encode elsewhere if you truly need MP3.

Fades are gain ramps, not filters

The fade-in slider multiplies the first milliseconds of the selection by a ramp from 0 to 1, and fade-out does the mirror at the end. That removes the click you get from cutting a waveform mid-slope; the spectral content in between stays untouched.

Frequently Asked Questions