Update app.py

app.py

@@ -20,7 +20,12 @@ pipe = pipeline(task="automatic-speech-recognition", model=model_repo_id, device
 def infer(
     audio
 ):
-    return pipe(audio[1], generate_kwargs={'language': 'chinese'})['text']
+    sampling_rate, wav = audio 
+    if wav.ndim > 1:
+        wav = wav.mean(axis=1)
+    wav = wav.astype(np.float32)
+    wav /= np.max(np.abs(wav))
+    return pipe({"sampling_rate": sampling_rate, "raw": wav}, generate_kwargs={'language': 'chinese'})['text']
 
 css = """
 #col-container {