bug in output

[werelord]

Not sure if this is the best place to report bugs..

Was testing out transcribing a podcast, which had some music at the end of the audio.. My encoding options:

whisper "audio.mp3" --model small.en --output_dir .\ --device cuda

The transcription was showing unicode characters for the stream, and an error was thrown trying to encode the data:

[1:08:14.740 --> 1:08:19.740]  ś
[1:08:19.740 --> 1:08:24.740]  ś
[1:08:24.740 --> 1:08:29.740]  ś
[1:08:29.740 --> 1:08:34.740]  ś
[1:08:34.740 --> 1:08:39.740]  ś
Traceback (most recent call last):
  File "C:\Python310\Scripts\whisper-script.py", line 33, in <module>
    sys.exit(load_entry_point('whisper==1.0', 'console_scripts', 'whisper')())
  File "C:\Python310\lib\site-packages\whisper\transcribe.py", line 291, in cli
    print(result["text"], file=txt)
  File "C:\Python310\lib\encodings\cp1252.py", line 19, in encode
    return codecs.charmap_encode(input,self.errors,encoding_table)[0]
UnicodeEncodeError: 'charmap' codec can't encode character '\u015b' in position 67044: character maps to <undefined>

[bcHelix]

I had this issue as well on Git Bash. See if setting export PYTHONIOENCODING=utf-8 or set if you're on CMD/PS works for you.