new

app.py

@@ -338,13 +338,24 @@ def extract_audio_from_video(video_path):
         audio_path
     ]
 
-    subprocess.run(
+    result = subprocess.run(
         command,
-        stdout=subprocess.DEVNULL,
-        stderr=subprocess.DEVNULL,
-        check=True
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE
     )
 
+    if result.returncode != 0:
+        error_msg = result.stderr.decode("utf-8", errors="replace")
+        print(f"[DEBUG] ffmpeg error: {error_msg}")
+        raise RuntimeError(
+            f"ffmpeg failed to extract audio: {error_msg[-200:]}"
+        )
+
+    if not os.path.exists(audio_path) or os.path.getsize(audio_path) == 0:
+        raise RuntimeError(
+            "ffmpeg produced no audio output — video may have no audio track"
+        )
+
     return audio_path
 
 # =========================================================
@@ -443,8 +454,25 @@ def predict_emotion(video_file):
 
     try:
 
+        # Handle both string paths and dict inputs from Gradio
+        if isinstance(video_file, dict):
+            video_path = video_file.get("video", video_file.get("name", None))
+        else:
+            video_path = video_file
+
+        if video_path is None or not os.path.exists(video_path):
+            return (
+                "⚠ VIDEO FILE NOT FOUND — The recorded/uploaded file could not be located.",
+                None,
+                ""
+            )
+
+        print(f"[DEBUG] Processing video: {video_path}")
+        print(f"[DEBUG] File exists: {os.path.exists(video_path)}")
+        print(f"[DEBUG] File size: {os.path.getsize(video_path)} bytes")
+
         audio_path = extract_audio_from_video(
-            video_file
+            video_path
         )
 
         transcribed_text = transcribe_audio(
@@ -460,7 +488,7 @@ def predict_emotion(video_file):
         ) = preprocess_inputs(
             audio_path,
             transcribed_text,
-            video_file
+            video_path
         )
 
         with torch.no_grad():
@@ -1522,6 +1550,7 @@ with gr.Blocks(
             video_input = gr.Video(
                 label="Video Input",
                 height=360,
+                sources=["upload", "webcam"],
                 elem_classes="black-label"
             )