Even out loudness and clean up sound in batch
PostLumina evens out loudness and cuts long silences across a whole folder of clips, on your own machine, when you ask Claude Code or Codex. It writes audio as M4A, MP3, WAV or FLAC. With speech-to-text on, your assistant can hear what is said in a clip before it acts.
Even out the loudness of these interviews and cut the long silences.
| Loudness | Normalised across the folder |
|---|---|
| Silences | Long silences trimmed |
| Also | Other audio processing, waveforms |
| Audio out | M4A, MP3, WAV, FLAC |
| Speech-to-text | On your machine with whisper.cpp: Small (fast) or Large v3 Turbo (accurate) |
| Not covered | Dialogue-gated loudness checks for Netflix-style delivery |
How to do it
- Add the folder to PostLumina as a project. It indexes every clip's format, duration and tags.
- Say the sentence to Claude Code or Codex, in your own words. Your assistant works out the steps; PostLumina runs them.
- Listen to the sample made from one or two clips. Once you like it, the whole folder runs, and the new files land in the project's PostLumina Output folder.
Questions
Which audio formats can it write?
M4A, MP3, WAV and FLAC.
Is my audio uploaded for speech-to-text?
No. Speech-to-text runs on your machine; no audio leaves it. When your assistant listens to a clip, it sees the transcript through your own Claude Code or Codex account.
Do I need to install anything else?
No. ffmpeg, Python and the speech engine come inside the app. The speech model is an optional download from within the app.