Merge pull request #319 from ziyu4huang/fix/audio-mixer-loudnorm-target

fix(audio_mixer): parameterize loudnorm LUFS target (§8 #1)

tests/tools/test_audio_mixer_loudnorm_target.py

@@ -0,0 +1,49 @@
+"""Regression for the loudnorm LUFS target parameterization (REVIEW §8 #1).
+
+audio_mixer hard-coded loudnorm I=-16 (Apple Podcasts), but
+sound-design.md targets -14 for YouTube/TikTok/IG and
+edit_decisions.metadata.loudnorm_target is the declarative form. The
+mismatch meant the executed loudness silently defaulted to podcast
+levels regardless of the target platform. These tests pin the per-call
+target resolution without invoking ffmpeg.
+"""
+
+from __future__ import annotations
+
+import math
+
+from tools.audio.audio_mixer import AudioMixer
+
+
+def _filter(inputs: dict) -> str:
+    return AudioMixer._loudnorm_filter(inputs, "premix", "out")
+
+
+def test_default_target_is_podcast_minus_16():
+    f = _filter({})
+    assert "I=-16.0" in f or "I=-16" in f
+    assert f.startswith("[premix]loudnorm=")
+    assert f.endswith("[out]")
+
+
+def test_youtube_target_minus_14_is_honored():
+    f = _filter({"loudnorm_target": -14})
+    assert "I=-14.0" in f or "I=-14" in f
+
+
+def test_out_of_range_target_is_clamped():
+    # A nonsense value must not produce a malformed ffmpeg arg.
+    f = _filter({"loudnorm_target": 99})
+    # Clamped to the 0 LUFS ceiling.
+    assert "I=0.0" in f
+
+
+def test_non_numeric_target_falls_back_to_default():
+    f = _filter({"loudnorm_target": "not-a-number"})
+    assert "I=-16.0" in f or "I=-16" in f
+
+
+def test_schema_exposes_loudnorm_target_default():
+    prop = AudioMixer().input_schema["properties"]["loudnorm_target"]
+    assert prop["type"] == "number"
+    assert math.isclose(prop["default"], -16)

tools/audio/audio_mixer.py

@@ -131,6 +131,20 @@ class AudioMixer(BaseTool):
                 },
             },
             "normalize": {"type": "boolean", "default": True},
+            "loudnorm_target": {
+                "type": "number",
+                "default": -16,
+                "minimum": -40,
+                "maximum": 0,
+                "description": (
+                    "Integrated loudness target (LUFS) for the loudnorm filter when "
+                    "normalize=true. Default -16 (Apple Podcasts). Pass -14 for "
+                    "YouTube/TikTok/IG per sound-design.md. Matches the "
+                    "edit_decisions.metadata.loudnorm_target convention — directors "
+                    "should forward that field here so the executed loudness matches "
+                    "the platform the asset targets."
+                ),
+            },
             "video_path": {
                 "type": "string",
                 "description": (
@@ -180,6 +194,25 @@ class AudioMixer(BaseTool):
         "Listen to mixed output and verify speech clarity and music ducking",
     ]
 
+    @staticmethod
+    def _loudnorm_filter(inputs: dict[str, Any], in_label: str, out_label: str) -> str:
+        """Build a loudnorm filter graph edge honoring the per-call LUFS target.
+
+        The integrated loudness target (``I=``) was historically hard-coded to
+        -16 (podcast/Apple). sound-design.md targets -14 for YouTube/TikTok/IG,
+        and edit_decisions.metadata.loudnorm_target is the declarative form.
+        Forward that value (or pass loudnorm_target directly) so the executed
+        loudness matches the target platform instead of silently defaulting.
+        """
+        target = inputs.get("loudnorm_target", -16)
+        try:
+            target = float(target)
+        except (TypeError, ValueError):
+            target = -16.0
+        # Clamp to a sane loudness range to avoid malformed ffmpeg args.
+        target = max(-40.0, min(0.0, target))
+        return f"[{in_label}]loudnorm=I={target}:LRA=11:TP=-1.5[{out_label}]"
+
     def execute(self, inputs: dict[str, Any]) -> ToolResult:
         operation = inputs["operation"]
         start = time.time()
@@ -251,7 +284,7 @@ class AudioMixer(BaseTool):
         )
 
         if normalize:
-            filter_parts.append("[mixed]loudnorm=I=-16:LRA=11:TP=-1.5[out]")
+            filter_parts.append(self._loudnorm_filter(inputs, "mixed", "out"))
             out_label = "[out]"
         else:
             out_label = "[mixed]"
@@ -558,7 +591,7 @@ class AudioMixer(BaseTool):
 
         # Normalize
         if normalize:
-            filter_parts.append("[premix]loudnorm=I=-16:LRA=11:TP=-1.5[out]")
+            filter_parts.append(self._loudnorm_filter(inputs, "premix", "out"))
             out_label = "[out]"
         else:
             out_label = "[premix]"