Further to this, the ma_encoder API in miniaudio only supports output to WAV. If you're needing to output to some other format you'll need to use a separate library for that which I can't really help you with.

You'll want to initialize a capture device. See the simple_capture example - that should cover almost everything you need. In the config, set the sample rate and channel count to 0 and the format to ma_format_unknown (or alternatively just don't assign anything to them and leave them as their defaults). By doing that, miniaudio will use whatever format/channels/rate is reported by the driver.

After the device has been initialized, you can inspect the following members of the ma_device object:

ma_format format     = device.capture.format;
ma_uint32 channels   = device.capture.channels;
ma_uint32 sampleRate = device.sampleRate;

The data callback will give you the raw data. You can use ma_encoder to write it to a file, which is also demonstrated in that example. You would need to change that example just a little bit so that you initialize the encoding using the format/channels/rate I mentioned just above instead of the hardcoded ones in the example which means you'll need to initialize the decoder after the device rather than before it like in the example.

Hi, Thanks for your reply, I had known how to record and set specified device long ago. simple_capture.c records audios into a file directly, but I want to read every frame raw data in buffer, how to set( such as frame size) and read the buffer? I guess I should do something in data_callback(), but I don't know how, maybe you can show me some code to do following things: record audio into a buffer with 256 data frame size and 8 channels(total buffer size should be 256 * 8 = 2048), and every sample point multiple by 2. Thanks for your kind reply!

I'm not fully sure what you mean. Instead of outputting to a file, do you want to instead output the captured audio to an in-memory buffer? It's just data - you just copy it like any other data. The size of the data in bytes of the captured is just the number of PCM frames multiplied by the number of channels multiplied by the size of a sample (defined by the sample format).

What do you mean by frame size? Do you mean the number of frames that are processed each time by the data callback? That's just specified in the device config: deviceConfig.periosSizeInFrames = 256;.

I'm not fully sure what you mean. Instead of outputting to a file, do you want to instead output the captured audio to an in-memory buffer? It's just data - you just copy it like any other data. The size of the data in bytes of the captured is just the number of PCM frames multiplied by the number of channels multiplied by the size of a sample (defined by the sample format).

What do you mean by frame size? Do you mean the number of frames that are processed each time by the data callback? That's just specified in the device config: deviceConfig.periosSizeInFrames = 256;.

Yes, I want to instead output the captured audio to an in-memory buffer. so I can process these data in real-time and write processed data to file. but I dont know how to set this buffer.

Frame size is what you said, you got it.

You can calculate the size of a buffer like this:

sizeInBytes = deviceConfig.periodSizeInFrames * ma_get_bytes_per_frame(device.capture.format, device.capture.channels);

That's controlled by the encoder. It should be obvious by just reading the code. Just get rid of the code relating to the encoder. I can't give you advice on every tiny little detail in your program. What helps is to just consider the audio data like any other data. The size of the data can be calculated like I showed you in my earlier comment. Whenever miniaudio gives you a pointer to a buffer containing some audio data, it'll have a frame count associated with it which you can use to determine the size of the data. Just manipulate the data like you would any other data.