Error in Following Deepspeech Example with Microphone


I am new in using DeepSpeech and currently following this example: mic_vad_streaming . I am using a Raspberry Pi 4 Model B and successfully ran deepspeech with the pre-trained models.

In the example linked above, I am having trouble with installing the “requirements.txt” file. I receive the error “ERROR: Invalid requirement: ‘’ (from line 6 of requirements.txt)”.

The end goal of my work with DeepSpeech will be to have a dedicated Pi with hotword detection and this software to run commands.

Any help would be appreciated. Thank you!

A bit strange, but why don’t you install the requirements by hand? You are using a virtual environment? And check out how mycroft does it, as they provide a DeepSpeech integration and are Raspberry based.

I have troubles understanding your problem, since you have not shared meaningful output / error / command line.

It would be because of scipy but again, with no clear error, we can’t do anything. Scipy is provided prebuilt by piwheels projects.

Please understand those examples are provided as-is, are user-contributed and there is no guarantee.

Hi Olaf,

I managed to install everything needed and I am using a virtual environment. Thank you for suggesting myCroft; I will definitely look into that!


I managed to install the requirements; however, I am running into issues when running the python script:

"Traceback (most recent call last):
    File "", line 224, in <module>
    File "", line 159, in main
        ARGS.scorer = os.path.join(model_dir, ARGS.scorer)
    File "/usr/lib/python3.7/", line 149, in join
        genericpath._check_arg_types('join', a, *p)
    File "/usr/lib/python3.7/", line 149, in _check_arg_types
        (funcname, s._class_._name_)) from none
TYPEError: join() argument must be str or bytes, not 'NoneType'"

I am currently running through the code to see if I can solve the issue. I understand that these examples are user-contributed and I appreciate the replies. Thank you!

  1. Format your code/err msg.

  2. Read the error msg. It tells you that there is no scorer.

  3. Read the docs to know why you need a scorer.