Online Audio Recorder

The Audio Recorder captures your microphone through the browser's getUserMedia API and records with MediaRecorder using the audio/webm;codecs=opus profile (falling back to plain audio/webm if Opus isn't supported, checked via MediaRecorder.isTypeSupported), collecting data in 100 ms chunks (mediaRecorder.start(100)) so pause and resume work cleanly. While you record, an AnalyserNode with an fftSize of 256 drives a live 32-bar waveform (the first 32 frequency bins) plus a volume meter that turns green below 30%, yellow up to 70%, and red at or above 70% so you can watch your input level before a take. The Recording Settings panel lets you pick a specific input from the devices returned by enumerateDevices, arm a 3, 5, or 10 second countdown before capture starts, and set a maximum duration with a slider that runs from 0 (Unlimited) to 300 seconds in 10 second steps, after which recording stops automatically with a toast.

Because browsers can only record Opus-in-WebM natively (they cannot encode MP3 or WAV directly), the MP3, WAV, and OGG download options are produced by re-encoding the captured WebM in your browser with FFmpeg compiled to WebAssembly (the convertAudioFormat helper). That conversion runs a plain ffmpeg -i input output command, letting FFmpeg pick the default encoder and bitrate for the chosen container (libmp3lame for .mp3, libvorbis for .ogg, and pcm_s16le 16-bit PCM for .wav) rather than forcing a specific rate. If a recording is tiny (under 100 bytes) or the conversion throws or returns an empty file, the tool falls back to handing you the original WebM rather than an empty file, and every download filename is timestamped from the current ISO time. The optional trim step also runs through FFmpeg (the trimAudio helper): it re-encodes the selected start-to-end region back to an intermediate WebM with libopus at 128 kbps so it can still be converted to your chosen format on download, with the trim sliders stepping at 0.01 second resolution.

Everything from capture to encoding happens on-device; the audio never leaves your machine, and FFmpeg loads from local WebAssembly files under /ffmpeg (with an unpkg CDN fallback) rather than a server upload. One practical consequence of the WebM source is that MediaRecorder files frequently report an Infinity duration, so the player and trim range fall back to a duration computed from the recording's own start/pause timestamps (Date.now() minus start minus total paused time). The tool only records the microphone selected through getUserMedia, so it cannot capture system or tab audio without an OS-level loopback device, and a denied permission or missing microphone surfaces a specific NotAllowedError or NotFoundError toast instead of failing silently.