update

examples/wenet/toolbox_infer.py

@@ -41,6 +41,18 @@ def get_args():
 def main():
     args = get_args()
 
+    # audio convert
+    signal, sample_rate = librosa.load(args.filename, sr=args.sample_rate)
+    signal *= 32768.0
+    signal = np.array(signal, dtype=np.int16)
+
+    temp_file = temp_directory / "temp.wav"
+    wavfile.write(
+        temp_file.as_posix(),
+        rate=args.sample_rate,
+        data=signal
+    )
+
     # audio convert
     in_filename = Path(args.in_filename)
     out_filename = Path(tempfile.gettempdir()) / "asr" / in_filename.name
@@ -69,7 +81,7 @@ def main():
     )
 
     text = decode.decode_by_recognizer(recognizer=recognizer,
-                                       filename=out_filename.as_posix(),
+                                       filename=temp_file,
                                        )
     print("text: {}".format(text))
     return