# Podcast talking-head cleanup

Clean up a single-camera talking-head recording.

## Requirements
- DaVinci Resolve with Resolve MCP
- whisper (or whisper.cpp) and ffmpeg installed

## Steps
1. Extract audio with ffmpeg, transcribe with whisper using word timestamps.
2. Build a cut list: remove silences longer than 0.7s and filler words ("um", "uh", "like" when isolated). Keep 0.15s of padding around each kept segment.
3. Apply the cut list in Resolve via MCP razor-and-delete on the timeline.
4. Alternate a 1.0x and 1.15x digital punch-in on each sentence boundary to hide jump cuts.
5. Normalize dialogue to -14 LUFS with a limiter at -1 dBTP.

## Verify
- Playback the first 60 seconds. No cut should clip a word.
- Check the loudness meter reads -14 LUFS integrated.