| name | Podcast cleanup |
| description | Clean up a recorded conversation: denoise voices, level volumes, duck music, and add gentle fades. |
Podcast cleanup
Goal: a talk recording that sounds clean and even, without changing its
content.
- Call
describe_project. Identify voice clips (clips on audio lanes, or
media clips whose audio carries the conversation) and any music clips.
- Tag roles so later steps and the UI know what is what:
set_audio_role with voiceover on each voice clip and music on
music clips.
- Denoise every voice clip:
set_denoise with denoise: true. Do not
denoise music; it smears.
- Level the conversation: voice clips should sit at volume 1.0; if one
speaker's clip is noticeably hot or quiet the user will say so. Adjust
only named clips with
set_clip_audio (e.g. volume 0.8), never guess.
- Duck music under speech: on each music clip that overlaps voice, set
set_clip_audio volume to ~0.25 and add 0.5 s fades in and out.
- Trim dead air at the ends: if the first or last clip starts or ends
with obvious silence the user mentioned,
trim_clip it tight, then add
a 0.3 s fade in on the first audible clip and a 0.5 s fade out on the
last (set_clip_audio).
- Report what changed per clip in one short list.