Test on lengths

app.py

@@ -72,6 +72,7 @@ from model import AVSEModule
 from config import sampling_rate
 import spaces
 import time
+import shutil
 
 # Load model once globally
 #ckpt_path = "ckpts/ep215_0906.oat.ckpt"
@@ -252,6 +253,18 @@ def extract_faces(video_file):
 
     enhanced_audio_path = run_avse_inference(output_path, audio_path)
 
+    # For API:
+    static_dir = "./file_outputs"
+    os.makedirs(static_dir, exist_ok=True)
+
+    # Move/copy to static
+    face_video_out = os.path.join(static_dir, "face_only_video.mp4")
+    enhanced_audio_out = os.path.join(static_dir, "enhanced_audio.wav")
+    shutil.copy(output_path, face_video_out)
+    shutil.copy(enhanced_audio_path, enhanced_audio_out)
+
+    return face_video_out, enhanced_audio_out
+
 
     return output_path, enhanced_audio_path
     #return output_path, audio_path