feat: MuseTalk V15 lip sync skill

Add AI lip sync skill powered by MuseTalk V15 — synchronizes lip
movements to match provided audio. Video+audio and image+audio inputs,
multi-face support, batch inference. Subprocess isolation for CUDA
memory safety. Path traversal protection.

Author: AEmotionStudio

skills/category/audio.py

@@ -806,3 +806,51 @@ def register_skills(registry: SkillRegistry) -> None:
             "non_commercial", "cc_by_nc",
         ],
     ))
+
+    # Lip sync — AI-powered lip synchronization (MuseTalk)
+    registry.register(Skill(
+        name="lip_sync",
+        category=SkillCategory.AUDIO,
+        description=(
+            "AI lip sync: synchronize a face's lip movements to match "
+            "provided audio using MuseTalk. Works with both video and image "
+            "inputs. Supports multiple faces. Requires an audio_path "
+            "pointing to the speech/voice audio file."
+        ),
+        parameters=[
+            SkillParameter(
+                name="audio_path",
+                type=ParameterType.STRING,
+                description="Path to audio file for lip sync (speech/voice)",
+                required=True,
+            ),
+            SkillParameter(
+                name="face_index",
+                type=ParameterType.INT,
+                description="Which face to sync: -1 = all faces, 0 = first face, 1 = second, etc.",
+                required=False,
+                default=-1,
+                min_value=-1,
+                max_value=9,
+            ),
+            SkillParameter(
+                name="batch_size",
+                type=ParameterType.INT,
+                description="Frames per inference batch (lower = less VRAM)",
+                required=False,
+                default=8,
+                min_value=1,
+                max_value=32,
+            ),
+        ],
+        examples=[
+            "lip_sync:audio_path=/path/to/speech.wav - Sync lips to speech audio",
+            "lip_sync:audio_path=/path/to/voice.mp3,face_index=0 - Sync first face only",
+            "lip_sync:audio_path=/path/to/audio.wav,batch_size=4 - Lower VRAM usage",
+        ],
+        tags=[
+            "lip", "sync", "lipsync", "dub", "dubbing", "talking", "head",
+            "face", "voice", "speech", "musetalk", "ai",
+        ],
+    ))
+

skills/composer.py

@@ -152,6 +152,14 @@ class SkillComposer:
         "synthesize_audio": "generate_audio",
         "generate_sound": "generate_audio",
         "add_sound": "generate_audio",
+        # MuseTalk — Lip Sync
+        "lipsync": "lip_sync",
+        "dub": "lip_sync",
+        "dubbing": "lip_sync",
+        "sync_lips": "lip_sync",
+        "talking_head": "lip_sync",
+        "lip_dub": "lip_sync",
+        "voice_sync": "lip_sync",
     }
 
     def __init__(self, registry: Optional[SkillRegistry] = None):
@@ -1240,6 +1248,8 @@ def _get_dispatch() -> dict:
         _f_auto_mask,
         # generate audio (MMAudio)
         _f_generate_audio,
+        # lip sync (MuseTalk)
+        _f_lip_sync,
         # presets
         _f_fade_to_black, _f_fade_to_white, _f_flash,
         _f_spin, _f_shake, _f_pulse, _f_bounce, _f_drift,
@@ -1419,6 +1429,15 @@ def _get_dispatch() -> dict:
         "synthesize_audio": _f_generate_audio,
         "generate_sound": _f_generate_audio,
         "add_sound": _f_generate_audio,
+        # MuseTalk — Lip Sync
+        "lip_sync": _f_lip_sync,
+        "lipsync": _f_lip_sync,
+        "dub": _f_lip_sync,
+        "dubbing": _f_lip_sync,
+        "sync_lips": _f_lip_sync,
+        "talking_head": _f_lip_sync,
+        "lip_dub": _f_lip_sync,
+        "voice_sync": _f_lip_sync,
         # Presets / Transitions
         "fade_to_black": _f_fade_to_black,
         "fade_to_white": _f_fade_to_white,

skills/handlers/__init__.py

@@ -15,6 +15,10 @@
     _f_generate_audio,
 )
 
+from .lip_sync import (  # noqa: F401
+    _f_lip_sync,
+)
+
 from .multi_input import (  # noqa: F401
     _f_grid,
     _f_slideshow,