Command-line voiceovers for automated pipelines

Built for narrated courses and lectures

Each subtitle cue becomes a separate audio segment. Revoicer places those segments on the WebVTT timeline, handles overlaps, trims trailing silence, and reuses cached segments when only part of a script changes.

A four-command lecture refresh

The packaged executables can be combined in a script for courses, lectures, and training libraries. A typical refresh looks like this:

1. Transcribe the original talk. Run transcribe on the lecture recording to generate an editable subtitle file.
2. Polish the script. Use transcript2vtt to turn the transcript into WebVTT, tweak the copy in your editor, and commit it alongside your slides.
3. Regenerate narration. Invoke voicer in your CI job to synthesize clean audio that matches the timing in the VTT file.
4. Distribute automatically. Optionally pipe the output through audiosplit to create per-slide clips before pushing the files to your LMS or video editor.

Everything runs headlessly, so you can schedule updates or trigger them from pull requests without manual studio time.