cut-silencelisted

## What this skill does Automatically cuts pauses and silence from talking-head videos using `auto-editor`, then normalizes the audio loudness with `ffmpeg-normalize`. - **Cost:** $0 in default mode — the tools are free, open source and fully offline - **Quality:** keeps the original video quality (no video re-encode during normalization, `-c:v copy`) - **Audio:** linear EBU R128 loudness normalization (no dynamic compression, no effects) — raises quiet audio and lowers loud audio to the same perceived level, without ever clipping - **Output:** writes the edited video to `<OUTPUT>/` (see convention below), never overwrites the original **Optional `--fillers` mode:** also removes filler words ("né", "tá", parasitic "tipo") and stutters. Costs a few cents of API (OpenRouter) because it has to transcribe, and it is **Brazilian Portuguese only** — the judgment prompt is written in PT-BR and calibrated on it. **Never runs unless the user asks.** See "Step 5b". --- ## Path convention This skill is portable: it works copied to any machine, without depending on any project's structure. | Marker | Means | |---|---| | `<OUTPUT>` | `<folder containing the video>/edited` — created if missing. **Never** the original's folder. | | `<SKILL-FOLDER>` | the folder this `SKILL.md` lives in. The script is at `<SKILL-FOLDER>/scripts/remove-fillers.py` | If the user asks for a different destination, use theirs. In batch mode, keep every video in the same `<OUTPUT>`. **Shell.** The command