How this recorder works

The full path, step by step

1. You press Record. The page calls navigator.mediaDevices.getUserMedia({ audio }), which triggers your browser's permission prompt.
2. The browser returns a MediaStream. This is a live in-memory handle to the microphone, not a file.
3. The stream is connected to an AudioContext, and an AnalyserNode reads it to draw the waveform and level meter.
4. For WAV and MP3, a processing node collects raw float samples into an array as you speak. For native output, MediaRecorder encodes the stream directly.
5. On Stop, the samples are converted — to 16-bit PCM with a RIFF header for WAV, or through a JavaScript LAME encoder for MP3.
6. The result becomes a Blob, and URL.createObjectURL() produces a blob: URL pointing at your own memory.
7. That URL is attached to a download link. Clicking it writes the file to your disk.

No step involves the network. The blob: URL scheme is specifically a reference to in-memory data in your browser, not a remote address.

Verifying it

• Network tab. Open developer tools, select Network, record and stop. No outbound request carrying audio appears.
• Read the source. The recorder is a single unminified JavaScript file at /recorder.js. It contains no fetch, XMLHttpRequest or WebSocket call. Search it and confirm.
• Test offline. Load the page, disconnect from the network, then record and download. It works, because none of it needs a connection.
• Check the download link. Right-click Download and copy the address — it begins blob:, which is in-memory data, not a server URL.

The offline test is the most convincing. A tool that requires an upload cannot function without a network; this one is unaffected.

What each format does technically

What the site loads

Everything is served from this domain. Fonts are self-hosted rather than pulled from a font CDN, so no third party learns that you visited. The MP3 encoder is served from this origin and loaded only when you select MP3. There are no third-party scripts on recorder pages.

The limitations this design imposes

Being local-only costs real capability, and it is worth stating plainly:

• No shareable links. Those require hosting the audio.
• No recovery. A closed tab means a lost recording.
• Memory-bound length. Recordings are held in RAM until downloaded.
• No background recording. The tab must stay visible.
• No server-side processing. Timeline editing, format conversion and noise reduction are not available.

These are consequences of the architecture, not oversights. A tool that fixed them would need to upload your audio.

• PrivacyWhat is and is not collected.
• How browser recording worksThe general technical background.
• Browser supportWhich browsers implement what.
• ContactTechnical questions welcome.

Questions