Update main.py

main.py

@@ -11,7 +11,7 @@ from moviepy.editor import (
     AudioFileClip,
     CompositeAudioClip,
     concatenate_audioclips,
-    vfx,  # Import vfx for looping
+    vfx,
 )
 import io
 from PIL import Image
@@ -25,20 +25,37 @@ from moviepy.config import change_settings
 change_settings({"VERBOSE": True})
 
 # Safe video clip loader
-def safe_video_clip(path):
+def safe_video_clip(path, target_width=1280, reference_aspect_ratio=None):
     try:
         clip = VideoFileClip(path)
         if clip.reader is None or clip.duration is None or clip.duration <= 0:
             clip.close()
             raise ValueError(f"Invalid or empty video file: {os.path.basename(path)}")
-        logger.debug(f"Loaded video: {path}, duration: {clip.duration}s")
-        # Resize without resample keyword
+        logger.debug(f"Loaded video: {path}, duration: {clip.duration}s, resolution: {clip.size}, aspect ratio: {clip.w / clip.h}")
+
+        # Get video aspect ratio
+        current_aspect_ratio = clip.w / clip.h
+
+        # Use reference aspect ratio if provided, else use current video's aspect ratio
+        target_aspect_ratio = reference_aspect_ratio if reference_aspect_ratio else current_aspect_ratio
+
+        # Check if aspect ratios are compatible (within a small tolerance)
+        if reference_aspect_ratio and abs(current_aspect_ratio - target_aspect_ratio) > 0.05:
+            logger.warning(f"Video {path} has different aspect ratio ({current_aspect_ratio:.2f}) than reference ({target_aspect_ratio:.2f}). Adjusting to match.")
+        
+        # Calculate target resolution to preserve aspect ratio
+        target_height = int(target_width / target_aspect_ratio)
+        # Ensure height is even (required by libx264 codec)
+        target_height = target_height if target_height % 2 == 0 else target_height + 1
+
         try:
-            clip = clip.resize((1280, 720))  # Use default resampling
+            clip = clip.resize((target_width, target_height))
+            logger.debug(f"Resized video {path} to {target_width}x{target_height}, aspect ratio: {target_width / target_height}")
         except Exception as e:
             logger.warning(f"Failed to resize video {path}: {e}")
-            # Continue without resizing if it fails
-        return clip
+            # Continue without resizing if it fails (may cause concatenation issues)
+        
+        return clip, target_aspect_ratio
     except Exception as e:
         raise ValueError(f"Failed to load video file '{os.path.basename(path)}': {str(e)}")
 
@@ -54,14 +71,17 @@ def safe_audio_clip(path):
     except Exception as e:
         raise ValueError(f"Failed to load audio file '{os.path.basename(path)}': {str(e)}")
 
-def merge_videos_and_audios(video_paths, audio_paths, output_path, temp_dir, orig_vol=1.0, music_vol=0.5, audio_mode="composite"):
+def merge_videos_and_audios(video_paths, audio_paths, output_path, temp_dir, orig_vol=1.0, music_vol=0.5, audio_mode="concatenate"):
     logger.debug(f"Merging {len(video_paths)} videos and {len(audio_paths)} audios to {output_path}")
     
     # Load video clips
     video_clips = []
+    reference_aspect_ratio = None
     for vp in video_paths:
         try:
-            clip = safe_video_clip(vp)
+            clip, aspect_ratio = safe_video_clip(vp, reference_aspect_ratio=reference_aspect_ratio)
+            if reference_aspect_ratio is None:
+                reference_aspect_ratio = aspect_ratio  # Set reference from first valid video
             video_clips.append(clip)
         except Exception as e:
             logger.warning(f"Skipping invalid video file {vp}: {e}")
@@ -81,7 +101,7 @@ def merge_videos_and_audios(video_paths, audio_paths, output_path, temp_dir, ori
 
     # Concatenate video clips
     final_video = concatenate_videoclips(video_clips, method="compose")
-    logger.debug(f"Concatenated video duration: {final_video.duration}s")
+    logger.debug(f"Concatenated video duration: {final_video.duration}s, resolution: {final_video.size}")
 
     # Handle original video audio
     original_audio = None
@@ -103,6 +123,9 @@ def merge_videos_and_audios(video_paths, audio_paths, output_path, temp_dir, ori
         if final_audio and final_audio.duration < final_video.duration:
             logger.debug(f"Looping audio to match video duration: {final_video.duration}s")
             final_audio = final_audio.fx(vfx.loop, duration=final_video.duration)
+        elif final_audio and final_audio.duration > final_video.duration:
+            logger.debug(f"Trimming audio to match video duration: {final_video.duration}s")
+            final_audio = final_audio.subclip(0, final_video.duration)
         
         if original_audio:
             final_audio = CompositeAudioClip([final_audio, original_audio])
@@ -148,7 +171,7 @@ async def merge_endpoint(
     files: list[UploadFile] = File(...),
     orig_vol: float = Form(1.0),
     music_vol: float = Form(0.5),
-    audio_mode: str = Form("composite"),
+    audio_mode: str = Form("concatenate"),
     background_tasks: BackgroundTasks = BackgroundTasks(),
 ):
     temp_dir = tempfile.mkdtemp()