Search code examples
caudioffmpegaaclibavcodec

Encode audio to aac with libavcodec


I'm using libavcodec (latest git as of 3/3/10) to encode raw pcm to aac (libfaac support enabled). I do this by calling avcodec_encode_audio repeatedly with codec_context->frame_size samples each time. The first four calls return successfully, but the fifth call never returns. When I use gdb to break, the stack is corrupt.

If I use audacity to export the pcm data to a .wav file, then I can use command-line ffmpeg to convert to aac without any issues, so I'm sure it's something I'm doing wrong.

I've written a small test program that duplicates my problem. It reads the test data from a file, which is available here: http://birdie.protoven.com/audio.pcm (~2 seconds of signed 16 bit LE pcm)

I can make it all work if I use FAAC directly, but the code would be a little cleaner if I could just use libavcodec, as I'm also encoding video, and writing both to an mp4.

ffmpeg version info:

FFmpeg version git-c280040, Copyright (c) 2000-2010 the FFmpeg developers
  built on Mar  3 2010 15:40:46 with gcc 4.4.1
  configuration: --enable-libfaac --enable-gpl --enable-nonfree --enable-version3 --enable-postproc --enable-pthreads --enable-debug=3 --enable-shared
  libavutil     50.10. 0 / 50.10. 0
  libavcodec    52.55. 0 / 52.55. 0
  libavformat   52.54. 0 / 52.54. 0
  libavdevice   52. 2. 0 / 52. 2. 0
  libswscale     0.10. 0 /  0.10. 0
  libpostproc   51. 2. 0 / 51. 2. 0

Is there something I'm not setting, or setting incorrectly in my codec context, maybe? Any help is greatly appreciated!

Here is my test code:

#include <stdio.h>
#include <libavcodec/avcodec.h>

void EncodeTest(int sampleRate, int channels, int audioBitrate,
    uint8_t *audioData, size_t audioSize)
{
    AVCodecContext  *audioCodec;
    AVCodec *codec;
    uint8_t *buf;
    int bufSize, frameBytes;

    avcodec_register_all();

    //Set up audio encoder
    codec = avcodec_find_encoder(CODEC_ID_AAC);
    if (codec == NULL) return;
    audioCodec = avcodec_alloc_context();
    audioCodec->bit_rate = audioBitrate;
    audioCodec->sample_fmt = SAMPLE_FMT_S16;
    audioCodec->sample_rate = sampleRate;
    audioCodec->channels = channels;
    audioCodec->profile = FF_PROFILE_AAC_MAIN;
    audioCodec->time_base = (AVRational){1, sampleRate};
    audioCodec->codec_type = CODEC_TYPE_AUDIO;
    if (avcodec_open(audioCodec, codec) < 0) return;

    bufSize = FF_MIN_BUFFER_SIZE * 10;
    buf = (uint8_t *)malloc(bufSize);
    if (buf == NULL) return;

    frameBytes = audioCodec->frame_size * audioCodec->channels * 2;
    while (audioSize >= frameBytes)
    {
        int packetSize;

        packetSize = avcodec_encode_audio(audioCodec, buf, bufSize, (short *)audioData);
        printf("encoder returned %d bytes of data\n", packetSize);
        audioData += frameBytes;
        audioSize -= frameBytes;
    }
}

int main()
{
    FILE *stream = fopen("audio.pcm", "rb");
    size_t size;
    uint8_t *buf;

    if (stream == NULL)
    {
        printf("Unable to open file\n");
        return 1;
    }

    fseek(stream, 0, SEEK_END);
    size = ftell(stream);
    fseek(stream, 0, SEEK_SET);
    buf = (uint8_t *)malloc(size);
    fread(buf, sizeof(uint8_t), size, stream);
    fclose(stream);

    EncodeTest(32000, 2, 448000, buf, size);
}

Solution

  • The problem seems to go away if the bitrate is less than 386000. Not sure why this is, as I can encode at bitrates higher than that using FAAC directly. But 128000 is good enough for my purposes, so I'm able to move forward.