output

app.py

@@ -97,15 +97,40 @@ with gr.Blocks() as demo:
             with gr.Column():
                 gr.HTML("""
                 <div>
-                    <span style="font-size: 20px;">🎬 Output Video</span><br>
+                    <span style="font-size: 20px;">🎬 Final Output Video</span><br>
                 </div>
                 """)
-                video_output = gr.Video(label="Output", height=512)
+                final_video = gr.Video(label="Final Output", height=512)
+
+        gr.HTML("<h3>Processing Steps</h3>")
+        with gr.Row():
+            with gr.Column():
+                gr.HTML('<p style="margin: 0; color: #666;">1. Original (Looped)</p>')
+                video_looped_output = gr.Video(label="Original Video", height=256)
+            with gr.Column():
+                gr.HTML('<p style="margin: 0; color: #666;">2. Face Cropped</p>')
+                face_cropped_output = gr.Video(label="Face Cropped", height=256)
+            with gr.Column():
+                gr.HTML('<p style="margin: 0; color: #666;">3. Lipsynced Face</p>')
+                lipsynced_face_output = gr.Video(label="Lipsynced Face", height=256)
+
+        with gr.Row():
+            with gr.Column():
+                gr.HTML(
+                    '<p style="margin: 0; color: #666;">4. Lipsynced Full (Blended)</p>'
+                )
+                lipsynced_full_output = gr.Video(label="Lipsynced Full", height=256)
 
     lipsync_only_btn.click(
         fn=lipsync_with_audio_target,
         inputs=[video_input, audio_input, session_state, crop_size_radio],
-        outputs=[video_output],
+        outputs=[
+            final_video,
+            video_looped_output,
+            face_cropped_output,
+            lipsynced_face_output,
+            lipsynced_full_output,
+        ],
     )
 
 

processing.py

@@ -50,6 +50,39 @@ def get_memory_usage():
     return f"RAM: {ram_used_gb:.2f}GB / {ram.total / (1024**3):.2f}GB ({ram_percent:.1f}%){gpu_info}"
 
 
+def validate_input(video_file, audio_file):
+    """Validate input files
+
+    Args:
+        video_file: Video input
+        audio_file: Audio input
+
+    Returns:
+        Tuple of (video_path, audio_path)
+    """
+    if video_file is None:
+        raise gr.Error("Please upload a video source.")
+    if audio_file is None:
+        raise gr.Error("Please upload a target audio.")
+
+    if isinstance(video_file, dict):
+        video_path = video_file.get("name") or video_file.get("path")
+    else:
+        video_path = video_file
+
+    if isinstance(audio_file, dict):
+        audio_path = audio_file.get("name") or audio_file.get("path")
+    else:
+        audio_path = audio_file
+
+    if video_path is None or not os.path.exists(video_path):
+        raise gr.Error("Could not read uploaded video file.")
+    if audio_path is None or not os.path.exists(audio_path):
+        raise gr.Error("Could not read uploaded audio file.")
+
+    return video_path, audio_path
+
+
 @spaces.GPU
 def process_lipsync_with_audio_target(
     video_file,
@@ -68,33 +101,15 @@ def process_lipsync_with_audio_target(
         progress: Progress tracking object
 
     Returns:
-        final_video_path
+        Tuple of (final_video, video_looped, face_cropped, lipsynced_face, lipsynced_full)
     """
     try:
-        if video_file is None:
-            raise gr.Error("Please upload a video source.")
-        if audio_file is None:
-            raise gr.Error("Please upload a target audio.")
+        video_path, audio_path = validate_input(video_file, audio_file)
 
         output_dir = setup_output_dir(session_id)
 
         logger.info(f"Memory at start: {get_memory_usage()}")
 
-        if isinstance(video_file, dict):
-            video_path = video_file.get("name") or video_file.get("path")
-        else:
-            video_path = video_file
-
-        if isinstance(audio_file, dict):
-            audio_path = audio_file.get("name") or audio_file.get("path")
-        else:
-            audio_path = audio_file
-
-        if video_path is None or not os.path.exists(video_path):
-            raise gr.Error("Could not read uploaded video file.")
-        if audio_path is None or not os.path.exists(audio_path):
-            raise gr.Error("Could not read uploaded audio file.")
-
         audio_duration = get_audio_duration(audio_path)
 
         progress(0.1, desc="🎬 Đang chuẩn bị video...")
@@ -158,15 +173,6 @@ def process_lipsync_with_audio_target(
         gc.collect()
         logger.info(f"Memory after lipsync: {get_memory_usage()}")
 
-        progress(0.8, desc="🔀 Đang ghép video...")
-        with timer("Blending face into original"):
-            lipsynced_full = blend_face_into_original(
-                video_looped, lipsynced_face, face_bbox, output_dir
-            )
-
-        gc.collect()
-        logger.info(f"Memory after lipsync: {get_memory_usage()}")
-
         progress(0.8, desc="🔀 Đang ghép video...")
         with timer("Blending face into original"):
             lipsynced_full = blend_face_into_original(
@@ -182,7 +188,7 @@ def process_lipsync_with_audio_target(
         progress(1.0, desc="✅ Hoàn tất!")
         logger.info(f"Memory at end: {get_memory_usage()}")
 
-        return final_video
+        return final_video, video_looped, face_cropped, lipsynced_face, lipsynced_full
 
     except Exception as e:
         print(f"ERROR in process_lipsync_with_audio_target: {e}")
@@ -197,7 +203,11 @@ def lipsync_with_audio_target(
     crop_size=256,
     progress=gr.Progress(track_tqdm=True),
 ):
-    """Wrapper for Gradio: Lipsync video source with audio target (English only)"""
+    """Wrapper for Gradio: Lipsync video source with audio target (English only)
+
+    Returns:
+        Tuple of (final_video, video_looped, face_cropped, lipsynced_face, lipsynced_full)
+    """
     if video_file is None:
         raise gr.Error("Please upload a video source.")
     if audio_file is None: