Add fade-in and fade-out effects to audio files with adjustable duration.
Load an audio file, choose a fade-in length, a fade-out length, or both, and download a new copy whose gain ramps from silence up to full level at the start and back down to silence at the end. Processing happens entirely in the browser, so nothing about the recording leaves the machine. The thing people get wrong first: the fades are anchored to the file's boundaries, not to where the sound actually starts and ends. A file with trailing silence gets a fade-out that happens partly inside that silence, and the audible content can still end with a click.
A concrete input and expected output from the current implementation.
Input
File: interview.mp3, duration 30 s. Fade-in: 2 s. Fade-out: 3 s.
Expected output
A 30-second file whose gain starts at zero, rises to full level by the 2-second mark, stays at full level from 2 to 27 seconds, and falls back to zero by the 30-second mark. Length, sample rate, and channel count are unchanged.
A fade is a gain ramp anchored to the file's boundaries: the fade-in occupies the first 2 seconds (0-2 s) and the fade-out the last 3 (27-30 s), leaving 25 seconds of unmodified audio between them.
A fade is a per-sample volume curve applied to the whole file. Every sample is multiplied by a gain that starts at zero, rises to full level across the fade-in duration, holds there, then falls back to zero across the fade-out duration. Because this only scales amplitudes, the output keeps the input's exact length, sample rate, and channel layout - nothing is cut, shifted, or resampled.
The shape of the ramp between zero and full matters to the ear. A linear ramp is simplest to compute, but a linear fade-out sounds as if it drops quickly and then lingers; an equal-power (cosine) curve keeps perceived loudness steadier. Whichever curve is used, the output is a new audio file, so a lossy input such as MP3 must be decoded, processed, and re-encoded - a second lossy generation.
Because the fade is anchored to the end of the file, not to the end of the sound. Any trailing silence - a pause after the last word, long room tone - sits between the content and the end of the file, so the gain starts falling while the audio is already quiet. Trim the tail before fading, or shorten the fade to fit the content.
The ramp's shape, not the fade itself, controls how it sounds. A linear gain ramp drops quickly in perceived loudness and then lingers, which can read as a dip; an equal-power (cosine) curve keeps perceived level steadier. If a fade sounds lumpy, it may be the curve the tool uses - there is usually nothing wrong with your file.
The example and behavioral notes were checked against the browser implementation. Standards and primary references below define the relevant format, formula, or platform behavior.
Convert CSV data into structured JSON arrays or objects with automatic type detection.
Transform JSON arrays and objects into downloadable CSV spreadsheet format.
Convert YAML configuration files to JSON format with validation and formatting.
Parse XML documents and convert them to equivalent JSON structure with attributes.
Convert tab-separated values to comma-separated format with proper escaping.
Convert numbers between binary, octal, decimal, and hexadecimal bases.