Update app.py

app.py

@@ -30,14 +30,29 @@ os.chmod('ffmpeg', st.st_mode | stat.S_IEXEC)
 model_size = "small"
 model = WhisperModel(model_size, device="cuda", compute_type="int8")
 
+def check_for_faces(video_path):
+    face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
+    cap = cv2.VideoCapture(video_path)
+
+    while True:
+        ret, frame = cap.read()
+        if not ret:
+            break
+
+        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+        faces = face_cascade.detectMultiScale(gray, 1.1, 4)
+
+        if len(faces) > 0:
+            return True
+
+    return False
+    
 def process_video(radio, video, target_language):
     if target_language is None:
         return gr.Interface.Warnings("Please select a Target Language for Dubbing.")
         
     run_uuid = uuid.uuid4().hex[:6]
-    
     output_filename = f"{run_uuid}_resized_video.mp4"
-    #ffmpeg.input(video).output(output_filename, vf='scale=-1:720:force_original_aspect_ratio=decrease').run()
     ffmpeg.input(video).output(output_filename, vf='scale=-2:720').run()
 
     video_path = output_filename
@@ -45,6 +60,14 @@ def process_video(radio, video, target_language):
     if not os.path.exists(video_path):
         return f"Error: {video_path} does not exist."
 
+    # Move the duration check here
+    video_info = ffmpeg.probe(video_path)
+    video_duration = float(video_info['streams'][0]['duration'])
+
+    if video_duration > 60:
+        os.remove(video_path)  # Delete the resized video
+        return gr.Interface.Warnings("Video duration exceeds 1 minute. Please upload a shorter video.")
+
     ffmpeg.input(video_path).output(f"{run_uuid}_output_audio.wav", acodec='pcm_s24le', ar=48000, map='a').run()
 
     #y, sr = sf.read(f"{run_uuid}_output_audio.wav")
@@ -87,8 +110,15 @@ def process_video(radio, video, target_language):
 
     video_path_fix = video_path
 
-    cmd = f"python Wav2Lip/inference.py --checkpoint_path 'Wav2Lip/checkpoints/wav2lip_gan.pth' --face {shlex.quote(video_path_fix)} --audio '{run_uuid}_output_synth.wav' --pads {pad_top} {pad_bottom} {pad_left} {pad_right} --resize_factor {rescaleFactor} --nosmooth --outfile '{run_uuid}_output_video.mp4'"
-    subprocess.run(cmd, shell=True)
+    has_face = check_for_faces(video_path)
+
+    if has_face:
+        cmd = f"python Wav2Lip/inference.py --checkpoint_path 'Wav2Lip/checkpoints/wav2lip_gan.pth' --face {shlex.quote(video_path)} --audio '{run_uuid}_output_synth.wav' --pads {pad_top} {pad_bottom} {pad_left} {pad_right} --resize_factor {rescaleFactor} --nosmooth --outfile '{run_uuid}_output_video.mp4'"
+        subprocess.run(cmd, shell=True)
+    else:
+        # Merge audio with the original video without running Wav2Lip
+        cmd = f"ffmpeg -i {video_path} -i {run_uuid}_output_synth.wav -c:v copy -c:a aac -strict experimental {run_uuid}_output_video.mp4"
+        subprocess.run(cmd, shell=True)
 
     if not os.path.exists(f"{run_uuid}_output_video.mp4"):
         raise FileNotFoundError(f"Error: {run_uuid}_output_video.mp4 was not generated.")