mix-audiolisted

Mix narration and BGM with sidechain ducking, loudness normalization to broadcast targets, and per-segment level matching. Use to produce the master audio bed for final render.
huodebing-alt/Agentic-Headshot-Video-Studio · ★ 0 · AI & Automation · score 71

Install: claude install-skill huodebing-alt/Agentic-Headshot-Video-Studio

# Mix Audio / 音频混音 ## When to use / 何时使用 After voice and BGM tracks chosen. Before render-final. 旁白和 BGM 选定后、最终渲染前。 ## Inputs / 输入 - `workspace/voice/master.wav`. - `inputs/bgm/<track>.mp3`. - `--target-lufs` default -16 (YouTube), -14 (Spotify), -23 (broadcast). ## Output / 输出 - `workspace/audio/master.wav`. - `workspace/audio/mix.report.json` with LUFS, TP, LRA. ## Steps / 步骤 1. Loop or trim BGM to total duration. 2. Apply sidechain ducking (`sidechaincompress`) keyed to voice envelope: threshold -28 dB, ratio 4:1, attack 5 ms, release 250 ms. 3. Sum tracks; loudnorm two-pass to `--target-lufs`. 4. Measure LUFS/TP/LRA; write report. ## Examples / 示例 ``` ffmpeg -i voice.wav -i bgm.mp3 -filter_complex "[1:a][0:a]sidechaincompress...; loudnorm=I=-16" master.wav ``` ## Failure recovery / 失败回退 - If BGM file missing, output voice-only at target LUFS. - If loudnorm clipping warnings, retry with TP=-2.