update

examples/wenet/toolbox_infer.py

@@ -54,13 +54,13 @@ def main():
     )
 
     # audio convert
-    in_filename = Path(args.in_filename)
-    out_filename = Path(tempfile.gettempdir()) / "asr" / in_filename.name
-    out_filename.parent.mkdir(parents=True, exist_ok=True)
-
-    audio_convert(in_filename=in_filename.as_posix(),
-                  out_filename=out_filename.as_posix(),
-                  )
+    # in_filename = Path(args.in_filename)
+    # out_filename = Path(tempfile.gettempdir()) / "asr" / in_filename.name
+    # out_filename.parent.mkdir(parents=True, exist_ok=True)
+    #
+    # audio_convert(in_filename=in_filename.as_posix(),
+    #               out_filename=out_filename.as_posix(),
+    #               )
 
     # load recognizer
     m_dict = models.model_map["Chinese"][0]
@@ -80,10 +80,21 @@ def main():
         num_active_paths=2,
     )
 
-    text = decode.decode_by_recognizer(recognizer=recognizer,
-                                       filename=temp_file.as_posix(),
-                                       )
+    s = recognizer.create_stream()
+
+    s.accept_wave_file(
+        temp_file.as_posix()
+    )
+    recognizer.decode_stream(s)
+
+    text = s.result.text.strip()
+    text = text.lower()
     print("text: {}".format(text))
+
+    # text = decode.decode_by_recognizer(recognizer=recognizer,
+    #                                    filename=temp_file.as_posix(),
+    #                                    )
+    # print("text: {}".format(text))
     return