C ++ FFmpeg音调问题

Question

I'm using swr_convert to lower/raise the pitch of incoming audio and store it in a .mp3. To change the pitch, I'm dividing the out sample rate by a factor. However, the resulting audio is slightly distorted when this factor is anything other than 1. Here's my conversion code:

...

// Set up resample context
swrContext = swr_alloc();
if (!swrContext)
    throw -15;

av_opt_set_int(swrContext, "in_channel_count", codecContext->channels, 0);
av_opt_set_int(swrContext, "in_channel_layout", codecContext->channel_layout, 0);
av_opt_set_int(swrContext, "in_sample_rate", codecContext->sample_rate, 0);
av_opt_set_sample_fmt(swrContext, "in_sample_fmt", codecContext->sample_fmt, 0);

av_opt_set_int(swrContext, "out_channel_count", STREAM_AUDIO_CHANNELS, 0);
av_opt_set_int(swrContext, "out_channel_layout", STREAM_AUDIO_CHANNEL_LAYOUT, 0);
av_opt_set_int(swrContext, "out_sample_rate", STREAM_AUDIO_SAMPLE_RATE / pitch, 0);
av_opt_set_sample_fmt(swrContext, "out_sample_fmt", STREAM_AUDIO_SAMPLE_FORMAT_GM, 0);

if (swr_init(swrContext))
    throw -16;

// Allocate re-usable frame
frameDecoded = av_frame_alloc();
if (!frameDecoded)
    throw -17;

frameDecoded->format = codecContext->sample_fmt;
frameDecoded->channel_layout = codecContext->channel_layout;
frameDecoded->channels = codecContext->channels;
frameDecoded->sample_rate = codecContext->sample_rate;

// Load frames
inPacket.data = NULL;
inPacket.size = 0;

int gotFrame, samples = 0;

while (av_read_frame(formatContext, &inPacket) >= 0) {

    if (inPacket.stream_index != streamId) 
        continue;

    if (avcodec_decode_audio4(codecContext, frameDecoded, &gotFrame, &inPacket) < 0)
        throw -18;

    if (!gotFrame)
        continue;

    // Begin conversion
    if (swr_convert(swrContext, NULL, 0, (const uint8_t **)frameDecoded->data, frameDecoded->nb_samples) < 0)
        throw -19;

    while (swr_get_out_samples(swrContext, 0) >= RAW_AUDIO_FRAME_SIZE) {

        // Allocate data
        uint8_t **convertedData = NULL;
        if (av_samples_alloc_array_and_samples(&convertedData, NULL, STREAM_AUDIO_CHANNELS, RAW_AUDIO_FRAME_SIZE, STREAM_AUDIO_SAMPLE_FORMAT_GM, 0) < 0)
            throw -20;

        // Convert
        if (swr_convert(swrContext, convertedData, RAW_AUDIO_FRAME_SIZE, NULL, 0) < 0)
            throw -21;

        // Calculate buffer size
        size_t bufferSize = av_samples_get_buffer_size(NULL, STREAM_AUDIO_CHANNELS, RAW_AUDIO_FRAME_SIZE, STREAM_AUDIO_SAMPLE_FORMAT_GM, 0);
        if (bufferSize < 0)
            throw -22;

        fwrite(convertedData[0], 1, bufferSize, outStream);
        av_free(convertedData);
    }
}

...

STREAM_AUDIO_SAMPLE_RATE is defined as 44100. Here's the entire program if it helps: http://pastebin.com/5akEwNg4

The program generates a .mp3 with 25 notes that decrease in pitch. Here's an example of the distortion: http://www.stuffbydavid.com/dl/30256478.mp3

Can you spot anything incorrect about my conversion, or is my method of changing the pitch incorrect? Is there another way?

Answer 1

Your call to swr_convert() with NULL as input flushes the internal queue and that (indirectly) causes the distortions (because after flushing you submit new input). You need to call swr_convert with a valid input and output buffer (both non-NULL) in a loop until the file has finished decoding, and only then flush the queue with NULL input at the very end.