How to make sure that data callback of build_output_stream runs faster than data callback of build_input_stream ?

[21pages]

Audio data is collected on one computer and transferred to another computer for playback. If the acquisition frequency is higher than the playback frequency, the delay will increase.
There is a StreamConfig which can set buffer_size and sample_rate, and there is a timeout on version 15.1, but it is unused on wasapi for windows.

 Default output stream config:
      SupportedStreamConfig { channels: 2, sample_rate: SampleRate(48000), buffer_size: Range { min: 0, max: 4294967295 }, sample_format: F32 }
    All supported output stream configs:
      1.1. SupportedStreamConfigRange { channels: 2, min_sample_rate: SampleRate(48000), max_sample_rate: SampleRate(48000), buffer_size: Range { min: 0, max: 4294967295 }, sample_format: F32 }

It seems that the only parameter I can adjust is buffer_size, but it doesn't help.

cpal/src/traits.rs

Line 153 in 17e7ab6

fn build_output_stream<T, D, E>(

Another way is to change the build_output_stream callback data length, now it is 1024.

[roderickvd]

When you are crossing different clock domains, you will experience drift. The "feedback" example shows how to use a buffer to account for most of this, at the cost of higher latency. More advanced approaches would use asynchronous resampling, e.g. with Rubato.