Create app.py

app.py

@@ -0,0 +1,46 @@
+import gradio as gr
+from speechbrain.pretrained import EncoderClassifier
+import subprocess
+import os
+import uuid
+import yt_dlp
+
+model = EncoderClassifier.from_hparams("Jzuluaga/accent-id-commonaccent_ecapa")
+
+def download_video_from_url(url):
+    out_path = f"video_{uuid.uuid4()}.mp4"
+    ydl_opts = {
+        'format': 'bestaudio/best',
+        'outtmpl': out_path,
+        'quiet': True,
+    }
+    with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+        ydl.download([url])
+    return out_path
+
+def extract_audio(video_file):
+    audio_path = "audio.wav"
+    cmd = [
+        "ffmpeg", "-i", video_file, "-vn",
+        "-acodec", "pcm_s16le", "-ac", "1", "-ar", "16000",
+        audio_path, "-y"
+    ]
+    subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    return audio_path
+
+
+def classify_accent(input_file_or_url):
+    # Check if it's a URL
+    if input_file_or_url.startswith("http"):
+        video_path = download_video_from_url(input_file_or_url)
+    else:
+        video_path = input_file_or_url
+
+    audio_path = extract_audio(video_path)
+    out_probs, top_prob, top_idx, label = model.classify_file(audio_path)
+
+    top_labels = model.hparams.label_encoder.decode_ndim(torch.topk(out_probs, 3).indices.squeeze())
+    confidences = torch.topk(out_probs, 3).values.squeeze().tolist()
+    result = "\n".join([f"{l}: {p*100:.2f}%" for l, p in zip(top_labels, confidences)])
+
+    return label[0], f"{top_prob.item()*100:.2f}%", result