Project
Roulette

Podcast Clip Cutter

Turns a timestamp and an episode into a shareable clip with captions.

IntermediateCommand lineMedia10–22hAI-buildable 4/5no setupworth shipping

What you're making

You hear a great two minutes in a podcast and want to share it. Give this the file and a rough timestamp and it cuts the clip, transcribes it, burns in captions, and hands you something you can post. The clever bit is it nudges the cut to the nearest silence so you never chop someone off mid-word.

An audio tool that extracts a range, transcribes it, and renders a captioned clip. Snapping cut points to silence is what separates a usable clip from an awkward one.

UI
Logic
Useful

Hosting · Free forever

Deploys free on npm, PyPI or GitHub Releases.

Nothing to host. Publishing to a package registry is free and permanent, and anyone can run it with a single command.

Publish to npm — one command and anyone can run it.

Core features · 7

  • Cuts a range from an audio file by timestamp
  • Snaps the cut points to the nearest silence
  • Transcribes the clip locally
  • Renders captions burned into a shareable video
  • Adds a waveform or cover art background
  • Batch mode for several clips from one episode
  • Leaves the source file untouched

The interesting part

  • Detecting silence reliably when there's background noise
  • Keeping caption timing aligned after the cut points move
  • Transcription and rendering are both slow, so progress reporting matters

Once it works

  • Suggest clip-worthy moments automatically
  • Add speaker labels to the captions
  • Export vertical and square formats

Suggested stack

typescriptnodeffmpegwhisper

Track it