Update app.py

app.py

@@ -1,25 +1,33 @@
-import gradio as gr
 import torch
 import librosa
 from speechbrain.inference.classifiers import EncoderClassifier
+from pydub import AudioSegment
+import gradio as gr
+import os
 
+# Load model only once
 classifier = EncoderClassifier.from_hparams(
     source="Jzuluaga/accent-id-commonaccent_ecapa",
     savedir="pretrained_models/accent-id-commonaccent_ecapa"
 )
 
-def classify_audio(audio_path):
-    # librosa can load audio from video files too
-    waveform, sr = librosa.load(audio_path, sr=16000, mono=True)
+def classify_accent(video):
+    # 'video' will already be a path to the uploaded file
+    audio = AudioSegment.from_file(video, format="mp4")
+    audio.export("output.wav", format="wav")
+
+    waveform, sr = librosa.load("output.wav", sr=16000, mono=True)
     waveform_tensor = torch.tensor(waveform).unsqueeze(0)
-    _, score, _, label = classifier.classify_batch(waveform_tensor)
-    return f"Predicted Accent: {label[0]}\nConfidence: {score.item():.4f}"
-
-iface = gr.Interface(
-    fn=classify_audio,
-    inputs=gr.Audio(type="filepath"),  # <-- allows upload of audio or video files
-    outputs="text",
-    title="Accent Identifier"
-)
 
-iface.launch()
+    prediction = classifier.classify_batch(waveform_tensor)
+    _, score, _, text_lab = prediction
+
+    return f"Accent: {text_lab[0]} (Confidence: {score.item():.2f})"
+
+
+iface = gr.Interface(fn=classify_accent,
+                    inputs=gr.Video(),
+                     outputs="text")
+
+if __name__ == "__main__":
+    iface.launch()