Update app.py

app.py

@@ -3,38 +3,38 @@ import os
 import torchaudio
 from speechbrain.pretrained import EncoderClassifier
 
-def accent_detect(video_file):
-    # Save uploaded video
-    if isinstance(video_file, tuple):
-        video_path = video_file[0]
-    else:
-        video_path = "uploaded_input.mp4"
-        with open(video_path, "wb") as f:
-            f.write(video_file.read())
+def accent_detect(video_path):
+    # video_path is a string file path provided by Gradio
+    # Extract audio from the video using ffmpeg
+    os.system(f"ffmpeg -y -i '{video_path}' -ar 16000 -ac 1 -vn audio.wav")
 
-    # Extract audio
-    os.system(f"ffmpeg -y -i {video_path} -ar 16000 -ac 1 -vn audio.wav")
+    # Check if audio.wav was created and is of reasonable size
     if not os.path.exists("audio.wav") or os.path.getsize("audio.wav") < 1000:
         return "Audio extraction failed. Please check your file."
 
-    # Classify accent
+    # Load the accent classification model
     accent_model = EncoderClassifier.from_hparams(
         source="speechbrain/lang-id-commonlanguage_ecapa",
         savedir="tmp_accent_model"
     )
+    # Load audio
     signal, fs = torchaudio.load("audio.wav")
     if signal.shape[0] > 1:
         signal = signal[0].unsqueeze(0)
+    # Predict accent
     prediction = accent_model.classify_batch(signal)
     pred_label = prediction[3][0]
     pred_scores = prediction[1][0]
     confidence = float(pred_scores.max()) * 100
-    explanation = f"Predicted Accent: {pred_label} ({confidence:.1f}%)\nThe model is {confidence:.0f}% confident this is a {pred_label} English accent."
+    explanation = (
+        f"Predicted Accent: {pred_label} ({confidence:.1f}%)\n"
+        f"The model is {confidence:.0f}% confident this is a {pred_label} English accent."
+    )
     return explanation
 
 demo = gr.Interface(
     fn=accent_detect,
-    inputs=gr.Video(type="filepath", label="Upload a Video File (MP4, WEBM, etc.)"),
+    inputs=gr.Video(label="Upload a Video File (MP4, WEBM, etc.)"),
     outputs="text",
     title="🗣️ English Accent Classifier (Gradio Demo)",
     description="Upload a short video clip of English speech. This tool predicts the English accent and confidence."