Dual Channel Opus Encoding

Question

I have been using the opus library for a while now, mostly using mono channel audio. However, I have been working on dual channel audio and have been trying to get it to work with opus. I initially thought it would be pretty straight forward, because I'd just be increasing the number of channels in the opus_encoder_create function, and increasing my frame_size in the opus_encode function. However, when I do encode the audio, all I get out is a jumbled mess, of what appears to be a summation of the two waveforms. The input is PCM data, interleaved by each sample (L1, R1, L2, R2, .. LN, RN). I am unsure how to proceed, any advice, or sample code would be appreciated.

So far, I have tried supplying the data as (L1, L2,.. LN, R1, R2, ... RN) and when I do it as such, the audio is intelligible, but the two channels are still corrupted, and I don't believe that this is the correct approach.

Here is a sample of how I am doing the conversion

#define NUM_CHANNELS 2
#define SAMPLE_RATE 8000
#define FRAME_SIZE 960
#define OUTPUT_FRAME_BYTES FRAME_SIZE / 8

int main(void) {
    int err;
    OpusEncoder *encoder;

    int16_t pcm_data[FRAME_SIZE * NUM_CHANNELS];
    uint8_t compressed_data[OUTPUT_FRAME_BYTES * NUM_CHANNELS];

    while (true) {
        /* Gather Dual Channel PCM Data from Microphone, and interleave
            pcm_data[0] = left channel sample 1
            pcm_data[1] = right channel sample 1
            pcm_data[2] = left channel sample 2
            pcm_data[3] = right channel sample 2
            ...
            pcm_data[FRAME_SIZE * NUM_CHANNELS - 2] = left channel sample FRAME_SIZE
            pcm_data[FRAME_SIZE * NUM_CHANNELS - 1] = right channel sample FRAME_SIZE
         */

        encoder =
            opus_encoder_create(8000, NUM_CHANNELS, OPUS_APPLICATION_RESTRICTED_LOWDELAY, &err);

        opus_encoder_ctl(encoder, OPUS_SET_COMPLEXITY(0));

        opus_encode(encoder, pcm_data, FRAME_SIZE, compressed_data,
                    OUTPUT_FRAME_BYTES * NUM_CHANNELS);

        // Write compressed data to file
        opus_encoder_destroy(encoder);
    }
    return 0;
}

Answer 1

I think you have incorrect FRAME_SIZE. Official documentation says:

To encode a frame, opus_encode() or opus_encode_float() must be called with exactly one frame (2.5, 5, 10, 20, 40 or 60 ms) of audio data

Frame size can be calculated as frameSize = frameLength * sampleRate / 1000 where frameLength is in ms. So you should have frame size from 20 (for 2.5 ms) to 480 (for 60 ms).

Another thing - your compressed_data size shouldn't depend on FRAME_SIZE, at the link above you can find suggested size for the output payload:

4000 bytes is recommended