Silence Chapters

Podcasts, lectures and meetings have a natural rhythm: a topic ends, someone takes a breath, the next topic begins. That rhythm leaves a fingerprint in the audio — stretches where the energy drops near the noise floor. Silence-based chaptering reads that fingerprint directly, which makes it the fastest way to rough-cut a long recording into navigable pieces. Paid tools like Klipa.ai will remove silences to tighten a recording, but they don't split the result into chapters, and the only true splitters are Python command-line tools you have to install and script yourself.

The catch is that "silence" is relative. A studio podcast has a near-zero noise floor, so a 1.5-second gap is unmistakable. A meeting recorded on a laptop in a café has constant background energy, so the same threshold catches nothing — or everything, if you push it too far. That is why this tool puts the threshold and minimum-pause sliders in front of you with a live pause map instead of guessing: you can see the amber regions land on real topic breaks and stop before they start fragmenting ordinary breaths. The merge control handles the leftovers, folding stray 3-second fragments back into their neighbours.

A few honest limits to plan around. This is signal processing, not transcription: it finds pauses, not topics, so a recording with no real gaps (back-to-back rapid speech) won't split well no matter the settings. Clips export as WAV because browsers can't write MP3 natively — the files are larger than MP3 but lossless and editable anywhere, and any free converter will shrink them in seconds. And the 60-minute / 500 MB cap exists because decoding audio into raw samples is the hungriest operation a browser can do; beyond that you risk a tab crash. For a two-hour recording, split the file first and process each half.