diff --git a/skills/website-to-hyperframes/scripts/assemble-index.mjs b/skills/website-to-hyperframes/scripts/assemble-index.mjs index a6652c5e4f..e506e65f98 100644 --- a/skills/website-to-hyperframes/scripts/assemble-index.mjs +++ b/skills/website-to-hyperframes/scripts/assemble-index.mjs @@ -183,11 +183,18 @@ const hasGlobalVoice = existsSync(join(hyperframesDir, globalVoicePath)); // ---------- BGM volume: duck under narration if any voice is present ---------- +// Values track references/background-music.md: under-VO sits at 0.4–0.6 (the +// example storyboard line uses ~0.45 — see background-music.md:90), pure-music +// videos sit near full at 0.7–0.9 to clear the encoder noise floor (line 31). +// 0.5 is the mid of the simplest under-VO hold-low band; 0.9 is the top of the +// pure-music band. Agents wanting per-section ducking embed the same id twice +// at different volumes (pattern 2 in background-music.md) — the assembler's +// single-bed BGM emit is the floor, not the ceiling. const hasPerSceneVoice = playOrder.some( ({ scene }) => scene.voicePath && existsSync(join(hyperframesDir, scene.voicePath)), ); const hasAnyVoice = hasGlobalVoice || (voiceMode === "per-scene" && hasPerSceneVoice); -const BGM_VOLUME = hasAnyVoice ? "0.8" : "0.9"; +const BGM_VOLUME = hasAnyVoice ? "0.5" : "0.9"; // ---------- build elements in track order ---------- const body = [];