appvoid / vosper Goto Github PK

Hi, it can customize to force to use only with specific language?? I probe changing the model but no luck...

I'm trying to run Vosper on my M1 Macbook Pro and am getting the following error after the first transcription runs:

- Testing, testing, one, two, three.
Traceback (most recent call last):
  File "/Users/jgoldin/Development/third_party/vosper/main.py", line 6, in <module>
    _input = listener.listen(mic, rec)
  File "/Users/jgoldin/Development/third_party/vosper/listener.py", line 36, in listen
    soundData = data.read(4096)
  File "/Users/jgoldin/miniconda3/envs/3.9/lib/python3.9/site-packages/pyaudio/__init__.py", line 570, in read
    return pa.read_stream(self._stream, num_frames,
OSError: [Errno -9981] Input overflowed

Here's a screencap:

https://user-images.githubusercontent.com/1571956/223549835-6bac1bb2-7306-46a4-a3c7-8722af73800a.mov

I had to initially change recording.py to get the script to start (was complaining about wrong channel count since its a mono mic):

    recording = sd.rec(int(secs * 48000), samplerate=48000, channels=1)

And added some prints:

recording = sd.rec(int(0 * 48000), samplerate=48000, channels=1)

def record (secs=4):
    print("[recorder.py] Recording...")
    global recording
    recording = sd.rec(int(secs * 48000), samplerate=48000, channels=1)

def stop():
    print("[recorder.py] Stopping recording!")
    global recording
    sd.stop()
    os.system('rm speaker.wav')
    write('speaker.wav', 48000, recording)

I've tried changing the 4096 value in recorder.py with no real change:

    soundData = data.read(4096)

I'm assuming it has to be something wonky with the sample rate and processing rate, but I don't have enough contextual knowledge here to know, and ChatGPT is only getting me so far haha.

Any thoughts?

Good day, I apologize for my lack of understanding i would be interested to know if this could be a possible solution you use with

ideasman42/nerd-dictation#74

Nerd-dictation is a great tool for enabling you to use voice in any text box. it's one downfall is that it doesn't have support for punctuation and correct grammar. i'm actually using it now to write this post without using my keyboard much. the incorporation of whisper with it's correct punctuation and grammar would be a great asset.

i look forward to your feedback