search

s0d3s / PyAudioWPatch

🐍 PyAudio | PortAudio fork with WASAPI loopback support 🔊 Record audio from speakers on Windows
Other
138 stars 8 forks source link

frames_per_buffer in the example pawp_record_wasapi_loopback.py #13

Closed c2727c closed 1 year ago

c2727c commented 1 year ago

Hello, I was trying out the example pawp_record_wasapi_loopback.py. However the recored audio was extremely accelerated. I managed to solve the issue by setting frames_per_buffer=pyaudio.get_sample_size(pyaudio.paInt16)*1024 instead of pyaudio.get_sample_size(pyaudio.paInt16).

Could the frames_per_buffer in the example be wrong?

s0d3s commented 1 year ago

Hello! The answer to your question is fifty-fifty.

Any value will not be erroneous, but there is a caveat: large values ​​of frames_per_buffer can contribute to the appearance of a delay (and use more memory), and on the other hand, smaller values ​​​​have a bad effect on performance (because the callback is called more often)

I would say that it is better to take a value of 512 or greater, but again, this is at the discretion of the developer.

You see audio acceleration for one of two reasons: the performance of the device is not enough (so some frames are skipped), or you get a single-channel recording and save it as a two-channel one.

Given that the increase of frames_per_buffer helped - you have the first case.

single frame size = sample width * number of channels (bytes)