Files
ai-podcast/postprod.py
tcpsyn cb5665bca8 Add broadcast polish features to postprod pipeline
New 13-step pipeline:
- De-essing (split-band sibilance compression)
- Breath reduction (detect + attenuate by -12dB)
- HPF integrated into denoise step (80Hz rumble cut)
- Stereo imaging (host center, caller slight right, music Haas widening)
- Silence trimming (head/tail dead air removal)
- Fade in/out (equal-power sine curve, 1.5s/3.0s defaults)
- Auto chapter detection from stem activity
- Episode metadata (ID3 tags: title, artist, album, track, artwork)

Every new feature has a --no-* flag to disable individually.
Revert this commit to restore previous 9-step pipeline.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-12 04:02:47 -07:00

33 KiB