Arjay Porlas

Internal tool

Automating the most repetitive hour of podcast editing

A desktop app that detects who is speaking and crops the frame to them, replacing manual camera-switching in CapCut.

Problem

Editing a two-camera podcast means watching the whole episode and cutting to whoever is talking, one switch at a time. On a one-hour episode that is hundreds of manual decisions, none of which are creative, and all of which have to happen before the interesting editing can start.

Context and constraints

Built for a podcast I work on. No deadline pressure and no client — which meant I could take the time to solve it properly instead of shipping a script that only I could run.

What I built

An Electron and React desktop app with a Python and FFmpeg backend. It detects the active speaker across the episode and applies the crop automatically, producing a switched edit that a human then reviews and adjusts rather than builds from nothing.

Before / after

Before: an editor scrubbing a full hour in CapCut, making every camera switch by hand. After: the switching pass is done on import, and the editor starts from a working cut.

Result

In regular use. On a typical one-hour episode it removes an estimated 60% of the manual camera-switching work — the speaker identification and the crop. That figure is my own estimate from using it, not a measured benchmark, and I would rather say so than round it into something that sounds firmer than it is.

What I would improve next

Time one episode with a stopwatch, before and after, and replace the estimate with a real number. It is ten minutes of work and I should have done it already.

Tools used

  • Electron
  • React
  • Python
  • FFmpeg

Artifacts

The editor with a 20-minute episode loaded: three detected speakers mapped to names and screen sides, 378 transcript segments, and the generated crop timeline above the waveform.
The editor with a 20-minute episode loaded: three detected speakers mapped to names and screen sides, 378 transcript segments, and the generated crop timeline above the waveform.

The reason this is on the site is not the stack. Nobody buying a workflow system cares that it is Electron.

It is here because it is a small, complete example of the thing I actually do: notice that a person is spending an hour a week on a decision that does not require judgment, and take that hour back — without taking the judgment away from them. The editor still reviews every cut. They just no longer make all of them from scratch.

Want this built for your team?

Book a pilot conversation