From 41d0a6d66d92cbf1452eaeeebfbd325c05153ccb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Miguel=20=C3=81ngel?= <miguel.sierra@heygen.com>
Date: Sat, 11 Jul 2026 20:56:05 +0000
Subject: [PATCH] fix(talking-head): preserve source audio

---
 skills-manifest.json                              |  4 ++--
 skills/talking-head-recut/SKILL.md                | 12 ++++++++++++
 skills/talking-head-recut/media-contract.test.mjs | 13 +++++++++++++
 3 files changed, 27 insertions(+), 2 deletions(-)
 create mode 100644 skills/talking-head-recut/media-contract.test.mjs

diff --git a/skills-manifest.json b/skills-manifest.json
index 9321089cb1..7bf0050652 100644
--- a/skills-manifest.json
+++ b/skills-manifest.json
@@ -74,8 +74,8 @@
       "files": 2
     },
     "talking-head-recut": {
-      "hash": "d5c51342625c9952",
-      "files": 27
+      "hash": "1c7d859d214f78ea",
+      "files": 28
     },
     "website-to-video": {
       "hash": "e8143700c50b7469",
diff --git a/skills/talking-head-recut/SKILL.md b/skills/talking-head-recut/SKILL.md
index 9d7f1160bf..65fcba6ab9 100644
--- a/skills/talking-head-recut/SKILL.md
+++ b/skills/talking-head-recut/SKILL.md
@@ -936,6 +936,15 @@ ffmpeg -y -i "$VIDEO_PATH" -c:v libx264 -crf 18 -g 30 -keyint_min 30 \
           data-track-index="1"
         ></video>
       </div>
+      <!-- Preserve the source program audio while the visual video stays muted. -->
+      <audio
+        id="source-audio"
+        src="input-video.mp4"
+        data-start="0"
+        data-duration="121.2"
+        data-track-index="10"
+        data-volume="1"
+      ></audio>
 
       <!-- Layer 2: each card-host sits at the bounds dictated by its layout. -->
       <!-- IMPORTANT: every card-host MUST carry BOTH "card-host" and "clip" classes. -->
@@ -1162,6 +1171,9 @@ PRODUCER_BROWSER_GPU_MODE=hardware npx hyperframes render public \
 ```
 
 `hyperframes render <dir>` reads `<dir>/index.html` and produces the MP4.
+The canonical composition keeps the visual `<video>` muted and mounts the same
+source as the root `#source-audio` track, so the rendered MP4 preserves the
+talking-head audio without a manual remux.
 The flag `PRODUCER_BROWSER_GPU_MODE=hardware` (or `--browser-gpu`) is
 strongly recommended on macOS — software-only Chrome rendering times out
 on most laptops.
diff --git a/skills/talking-head-recut/media-contract.test.mjs b/skills/talking-head-recut/media-contract.test.mjs
new file mode 100644
index 0000000000..45a19f4425
--- /dev/null
+++ b/skills/talking-head-recut/media-contract.test.mjs
@@ -0,0 +1,13 @@
+import assert from "node:assert/strict";
+import { readFileSync } from "node:fs";
+import test from "node:test";
+
+const skill = readFileSync(new URL("./SKILL.md", import.meta.url), "utf8");
+
+test("canonical composition preserves source audio as a root media track", () => {
+  assert.match(skill, /<video[\s\S]*?id="bg-video"[\s\S]*?muted[\s\S]*?<\/video>/);
+  assert.match(
+    skill,
+    /<\/div>\s*<!-- Preserve the source program audio[\s\S]*?<audio[\s\S]*?id="source-audio"[\s\S]*?src="input-video\.mp4"[\s\S]*?data-track-index="10"[\s\S]*?<\/audio>/,
+  );
+});