Update app.py

app.py

@@ -1,10 +1,9 @@
-
-import streamlit as st
+import gradio as gr
 from moviepy.editor import VideoFileClip
 from speechbrain.pretrained import EncoderClassifier
 import torchaudio
-import os
 from pytube import YouTube
+import os
 
 CLASSIFIER = "Jzuluaga/accent-id-commonaccent_xlsr-en-english"
 
@@ -13,39 +12,62 @@ def download_video(url):
     if "youtube.com" in url:
         yt = YouTube(url)
         stream = yt.streams.filter(progressive=True, file_extension='mp4').first()
-        return stream.download()
-    else:  # Direct download
-        # Implement custom download logic for non-YouTube links
-        return url  
+        video_path = stream.download()
+        return video_path
+    else:  # Direct download (assumes URL is direct mp4 link)
+        # Download file locally
+        import requests
+        local_filename = "temp_video.mp4"
+        with requests.get(url, stream=True) as r:
+            r.raise_for_status()
+            with open(local_filename, 'wb') as f:
+                for chunk in r.iter_content(chunk_size=8192):
+                    f.write(chunk)
+        return local_filename
 
 def extract_audio(video_path):
     clip = VideoFileClip(video_path)
     audio_path = "temp_audio.wav"
-    clip.audio.write_audiofile(audio_path)
+    clip.audio.write_audiofile(audio_path, logger=None)
+    clip.close()
     return audio_path
 
 def classify_accent(audio_path):
     classifier = EncoderClassifier.from_hparams(
         source=CLASSIFIER,
-        savedir="pretrained_models/accent_classifier"
+        savedir="pretrained_models/accent_classifier",
+        run_opts={"device":"cpu"}  # or "cuda" if GPU available
     )
     waveform, sample_rate = torchaudio.load(audio_path)
     prediction = classifier.classify_batch(waveform)
-    return prediction
+    # prediction format: (scores, probabilities, embeddings, predicted_labels)
+    predicted_accent = prediction[3][0]
+    confidence = prediction[1].exp().max().item() * 100
+    return predicted_accent, f"{confidence:.2f}%"
+
+def process_video(url):
+    try:
+        video_path = download_video(url)
+        audio_path = extract_audio(video_path)
+        accent, confidence = classify_accent(audio_path)
+    finally:
+        # Cleanup temp files if they exist
+        for f in [video_path, audio_path]:
+            if os.path.exists(f):
+                os.remove(f)
+    return accent, confidence
 
-def main():
-    st.title("REM Waste Accent Classifier")
-    url = st.text_input("Enter video URL:", "")
-    
-    if url:
-        with st.spinner("Processing..."):
-            video_path = download_video(url)
-            audio_path = extract_audio(video_path)
-            prediction = classify_accent(audio_path)
-            
-            st.success("Analysis Complete")
-            st.metric("Detected Accent", prediction[3][0])
-            st.metric("Confidence Score", f"{prediction[1].exp().max().item()*100:.2f}%")
+# Gradio interface
+iface = gr.Interface(
+    fn=process_video,
+    inputs=gr.Textbox(label="Enter Public Video URL (YouTube, Loom, direct MP4)"),
+    outputs=[
+        gr.Textbox(label="Detected Accent"),
+        gr.Textbox(label="Confidence Score")
+    ],
+    title="English Accent Classifier",
+    description="Paste a public video URL to detect the English accent and confidence score."
+)
 
 if __name__ == "__main__":
-    main()
+    iface.launch()