Create app.py

app.py

@@ -0,0 +1,33 @@
+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_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)
+
+    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()