add app.py

app.py

@@ -1,53 +1,62 @@
 import os
+import sys
+import types
 import uuid
-import gradio as gr
-from wyn_wav2lip.wav2lip import Wav2Lip
 
-MEDIA_DIR = "media"
-os.makedirs(MEDIA_DIR, exist_ok=True)
+# --------------------------------------------------------------------
+# Fix environment issues
+# --------------------------------------------------------------------
+# Fix OMP_NUM_THREADS error
+os.environ["OMP_NUM_THREADS"] = "1"
 
-wav2lip = Wav2Lip()
-wav2lip.setup()
+# Dummy google.colab for wyn_wav2lip
+google = types.ModuleType("google")
+colab = types.ModuleType("google.colab")
 
-def lipsync_func(image, audio):
-    # image: PIL.Image
-    # audio: tuple (sample_rate, data) or filepath depending on Gradio component
+class _DummyDrive:
+    def mount(self, *args, **kwargs):
+        print("google.colab.drive.mount() called (dummy).")
 
-    # Save inputs
-    image_path = os.path.join(MEDIA_DIR, f"{uuid.uuid4()}.png")
-    audio_path = os.path.join(MEDIA_DIR, f"{uuid.uuid4()}.wav")
+colab.drive = _DummyDrive()
+google.colab = colab
 
-    image.save(image_path)
+sys.modules["google"] = google
+sys.modules["google.colab"] = colab
 
-    import soundfile as sf
-    sr, data = audio  # if using gr.Audio(type="numpy")
-    sf.write(audio_path, data, sr)
+# --------------------------------------------------------------------
+# Imports
+# --------------------------------------------------------------------
+import gradio as gr
+from wyn_wav2lip.wav2lip import Wav2Lip
+import soundfile as sf
 
-    # Run Wav2Lip like we did before
-    # (Use the same logic as in the Colab example)
-    video_path = run_wav2lip(image_path, audio_path)
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+MEDIA_DIR = os.path.join(BASE_DIR, "media")
+os.makedirs(MEDIA_DIR, exist_ok=True)
+
+print("Initialising Wav2Lip...")
+wav2lip = Wav2Lip()
+wav2lip.setup()
+print("Wav2Lip ready.")
 
-    return video_path
 
-def run_wav2lip(image_path, audio_path):
-    # similar to the Colab sample before
+def run_wav2lip(image_path: str, audio_path: str) -> str:
+    """Run Wav2Lip and return absolute path to created video file."""
     existing_mp4 = {
         f for f in os.listdir(MEDIA_DIR)
         if f.lower().endswith(".mp4")
     }
 
-    import os
-    from os import getcwd, chdir
-    
-    old = getcwd()
-    chdir(MEDIA_DIR)
+    old_cwd = os.getcwd()
+    os.chdir(MEDIA_DIR)
     try:
+        # Wav2Lip expects filenames relative to current working dir
         wav2lip.run(
             video_file=os.path.basename(image_path),
             vocal_file=os.path.basename(audio_path),
         )
     finally:
-        chdir(old)
+        os.chdir(old_cwd)
 
     new_mp4 = [
         f for f in os.listdir(MEDIA_DIR)
@@ -60,16 +69,45 @@ def run_wav2lip(image_path, audio_path):
             if f.lower().endswith(".mp4")
         ]
         if not mp4_candidates:
-            raise RuntimeError("No video created")
+            raise RuntimeError("No MP4 created by Wav2Lip.")
         return max(mp4_candidates, key=os.path.getmtime)
+
     return os.path.join(MEDIA_DIR, new_mp4[0])
 
 
+def lipsync_func(image, audio):
+    """
+    image: PIL image (from gr.Image)
+    audio: (sr, data) tuple (from gr.Audio(type="numpy"))
+    """
+    if image is None or audio is None:
+        return None
+
+    # Save image
+    img_id = uuid.uuid4().hex
+    image_path = os.path.join(MEDIA_DIR, f"{img_id}.png")
+    image.save(image_path)
+
+    # Save audio
+    aud_id = uuid.uuid4().hex
+    audio_path = os.path.join(MEDIA_DIR, f"{aud_id}.wav")
+    sr, data = audio
+    sf.write(audio_path, data, sr)
+
+    # Run Wav2Lip
+    video_path = run_wav2lip(image_path, audio_path)
+    return video_path
+
+
 demo = gr.Interface(
     fn=lipsync_func,
-    inputs=[gr.Image(type="pil"), gr.Audio(type="numpy")],
-    outputs=gr.Video(),
-    title="Wav2Lip Lipsync Service"
+    inputs=[
+        gr.Image(type="pil", label="Teacher image"),
+        gr.Audio(type="numpy", label="Teacher audio (.wav)")
+    ],
+    outputs=gr.Video(label="Lip-synced video"),
+    title="Wav2Lip Lipsync Service",
+    description="Upload a static teacher image and a WAV audio. The model will generate a talking video."
 )
 
 if __name__ == "__main__":

requirements.txt

@@ -1,2 +1,7 @@
 wyn-wav2lip
 gradio
+soundfile
+gdown
+dlib-bin==19.24.2.post1
+opencv-python
+