Update app.py

app.py

@@ -191,8 +191,19 @@ def infer_video_depth(
             # Speichern wir den Originalnamen
             original_path = stitched_video_path
             
-            # Temporärer Pfad für das geloopte Video
-            temp_looped_path = os.path.join(output_dir, 'temp_looped_rgbd.mp4')
+            # Überprüfen wir, ob das Input-Video einen Audio-Stream hat
+            has_audio = False
+            check_audio_cmd = [
+                "ffmpeg",
+                "-i", input_video,
+                "-c", "copy",
+                "-f", "null",
+                "-"
+            ]
+            result = subprocess.run(check_audio_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+            stderr = result.stderr.decode('utf-8')
+            if "Audio" in stderr:
+                has_audio = True
             
             # Erstelle eine temporäre Textdatei für die stitched Videos
             concat_stitched_file_path = os.path.join(output_dir, 'concat_stitched_list.txt')
@@ -200,8 +211,11 @@ def infer_video_depth(
                 for _ in range(loop_factor):
                     f.write(f"file '{original_path}'\n")
             
-            # Verwende ffmpeg, um das stitched Video zu loopen
-            cmd = [
+            # Temporärer Pfad für das geloopte Video ohne Audio
+            temp_looped_path = os.path.join(output_dir, 'temp_looped_rgbd.mp4')
+            
+            # Verwende ffmpeg, um das Video zu loopen
+            concat_cmd = [
                 "ffmpeg",
                 "-y",
                 "-f", "concat",
@@ -210,61 +224,72 @@ def infer_video_depth(
                 "-c", "copy",
                 temp_looped_path
             ]
-            subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
-            
-            # Extrahiere den Audio-Track aus dem ursprünglichen Input-Video
-            audio_path = os.path.join(output_dir, 'extracted_audio.aac')
-            extract_audio_cmd = [
-                "ffmpeg",
-                "-y",
-                "-i", input_video,
-                "-vn", "-acodec", "copy",
-                audio_path
-            ]
-            subprocess.run(extract_audio_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
-            
-            # Erstelle eine Textdatei für das Audio-Looping
-            concat_audio_file_path = os.path.join(output_dir, 'concat_audio_list.txt')
-            with open(concat_audio_file_path, 'w') as f:
-                for _ in range(loop_factor):
-                    f.write(f"file '{audio_path}'\n")
-            
-            # Erstelle den geloopten Audio-Track
-            looped_audio_path = os.path.join(output_dir, 'looped_audio.aac')
-            audio_loop_cmd = [
-                "ffmpeg",
-                "-y",
-                "-f", "concat",
-                "-safe", "0",
-                "-i", concat_audio_file_path,
-                "-c", "copy",
-                looped_audio_path
-            ]
-            subprocess.run(audio_loop_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
-            
-            # Kombiniere das geloopte Video mit dem geloopten Audio
-            # Aber schreibe zurück in den originalen Pfad
-            final_cmd = [
-                "ffmpeg",
-                "-y",
-                "-i", temp_looped_path,
-                "-i", looped_audio_path,
-                "-c:v", "copy",
-                "-c:a", "aac",
-                "-map", "0:v:0",
-                "-map", "1:a:0",
-                "-shortest",
-                original_path  # Verwenden des originalen Pfads als Ziel
-            ]
-            subprocess.run(final_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+            subprocess.run(concat_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
             
-            # Wir verwenden weiterhin den originalen Pfad
-            # (stitched_video_path bleibt unverändert)
+            # Wenn Audio vorhanden ist, müssen wir es separat behandeln
+            if has_audio:
+                # Extrahiere den Audio-Track aus dem originalen Input-Video
+                # Dies ist die sauberste Quelle
+                audio_path = os.path.join(output_dir, 'extracted_audio.aac')
+                extract_audio_cmd = [
+                    "ffmpeg",
+                    "-y",
+                    "-i", input_video,  # Original Input-Video verwenden
+                    "-vn", "-acodec", "copy",
+                    audio_path
+                ]
+                subprocess.run(extract_audio_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+                
+                # Erstelle eine Textdatei für das Audio-Looping
+                concat_audio_file_path = os.path.join(output_dir, 'concat_audio_list.txt')
+                with open(concat_audio_file_path, 'w') as f:
+                    for _ in range(loop_factor):
+                        f.write(f"file '{audio_path}'\n")
+                
+                # Erstelle den geloopten Audio-Track
+                looped_audio_path = os.path.join(output_dir, 'looped_audio.aac')
+                audio_loop_cmd = [
+                    "ffmpeg",
+                    "-y",
+                    "-f", "concat",
+                    "-safe", "0",
+                    "-i", concat_audio_file_path,
+                    "-c", "copy",
+                    looped_audio_path
+                ]
+                subprocess.run(audio_loop_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+                
+                # Kombiniere das geloopte Video mit dem geloopten Audio
+                final_cmd = [
+                    "ffmpeg",
+                    "-y",
+                    "-i", temp_looped_path,
+                    "-i", looped_audio_path,
+                    "-c:v", "copy",
+                    "-c:a", "aac",
+                    "-map", "0:v:0",
+                    "-map", "1:a:0",
+                    "-shortest",
+                    original_path  # Verwenden des originalen Pfads als Ziel
+                ]
+                subprocess.run(final_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+            else:
+                # Wenn kein Audio vorhanden ist, einfach das Video umbenennen
+                os.replace(temp_looped_path, original_path)
             
             # Bereinige temporäre Dateien
-            for file_path in [concat_file_path, concat_stitched_file_path, concat_audio_file_path, audio_path, looped_audio_path, temp_looped_path]:
+            temp_files = [concat_file_path, concat_stitched_file_path]
+            if has_audio:
+                temp_files.extend([concat_audio_file_path, audio_path, looped_audio_path])
+            if os.path.exists(temp_looped_path):
+                temp_files.append(temp_looped_path)
+                
+            for file_path in temp_files:
                 if os.path.exists(file_path):
-                    os.remove(file_path)
+                    try:
+                        os.remove(file_path)
+                    except:
+                        pass
 
     gc.collect()
     torch.cuda.empty_cache()