← ClaudeAtlas

audio-productionlisted

Deterministic audio production execution Skill for the AI Video Production Ecosystem. Use it when a caller (normally video-production-agent) has already decided what to do to an audio file and needs it executed safely - gain, trim, cut, removal of explicitly given silence ranges, removal of filler words at caller-supplied word timestamps (e.g. transcription-skill's transcript) against a caller-chosen lexicon, fade in / out, EBU R128 loudness normalisation to a given target and true-peak ceiling, several delivery loudness versions of one master in one run (EBU R128 / ATSC A/85 profiles or stated targets, each measured), mixing two to eight tracks (optionally ducking one bed under the first track), stems (dialogue / music / effects as separate files of the programme's length that sum to the mix), mono / stereo / surround down-mix, FFT noise reduction, speech clean-up presets (ffmpeg-skill's --voice light / medium / strong), conversion to wav / flac / mp3 / m4a / aac / ogg / opus - as a typed operation graph wit
zobaerio/zobvia · ★ 1 · Data & Documents · score 74
Install: claude install-skill zobaerio/zobvia
> **Bundled with Zobvia** (extras/audio-production-skill/) — see the sub-skill parts of Zobvia's SKILL.md. Source: [kajisho5/audio-production-skill](https://github.com/kajisho5/audio-production-skill) v1.3.2 (main @ d179cac), MIT License (LICENSE in this folder). Bundled: SKILL.md, `src/`, `pyproject.toml`. Not bundled: docs, tests, brand. Runs via `PYTHONPATH=src python3 -m audio_production.cli ...` or after `pip install .` as `audio-production ...`; needs FFmpeg on PATH. # audio-production Machine interface: `audio-production run - --json` with a `audio-production/request@1` document on stdin; exactly one `audio-production/response@1` document on stdout. `skill --json` prints the contract, `doctor --json` the environment, `plan - --json` a dry run, `validate - --json` a schema check. Rules for a calling agent: 1. Measure first with media-analysis-skill (or ffmpeg-skill probe / loudness --measure-only); decide; then build a request. 2. Give explicit, typed parameters: `NORMALIZE` needs `target_lufs` and `true_peak_db`; each `LOUDNESS_SET` version names a profile or states both, and the set needs `tolerance_lufs`; `SILENCE_REMOVE` needs the ranges you decided to remove; `FILLER_REMOVE` needs word timestamps and a `lexicon`. This skill has no defaults for production values and never guesses. 3. Never send commands, argv, filter strings or executable paths: the request is rejected. 4. Keep outputs inside the workspace, never at an input path; set `overwrite: true` delibe