FFmpeg library: webm (vorbis) audio to aac conversion

Question

I have written a small program to convert webm (vorbis) audio to aac format, using FFmpeg libraries - C++ (on Windows using 32 bit Zeranoe FFmpeg builds). After writing this program, I find it is sometimes converting files as per expectation, and at other times, results in larger duration files, and audio playback is broken/awkward as well.

This code appears to be working fine for mp3, which also uses FLTP format (same as vorbis), so technically both look similar.

Please see below sample code I am using:

////////////////////////////////////////////////
#include "stdafx.h"

#include <iostream>
#include <fstream>

#include <string>
#include <vector>
#include <map>

#include <deque>
#include <queue>

#include <math.h>
#include <stdlib.h>
#include <stdio.h>
#include <conio.h>

extern "C"
{
#include "libavcodec/avcodec.h"
#include "libavformat/avformat.h"
#include "libavdevice/avdevice.h"
#include "libswscale/swscale.h"
#include "libavutil/dict.h"
#include "libavutil/error.h"
#include "libavutil/opt.h"
#include <libavutil/fifo.h>
#include <libavutil/imgutils.h>
#include <libavutil/samplefmt.h>
#include <libswresample/swresample.h>
}

AVFormatContext*    fmt_ctx= NULL;
int                    audio_stream_index = -1;
AVCodecContext *    codec_ctx_audio = NULL;
AVCodec*            codec_audio = NULL;
AVFrame*            decoded_frame = NULL;
uint8_t**            audio_dst_data = NULL;
int                    got_frame = 0;
int                    audiobufsize = 0;
AVPacket            input_packet;
int                    audio_dst_linesize = 0;
int                    audio_dst_bufsize = 0;
SwrContext *        swr = NULL;

AVOutputFormat *    output_format = NULL ;
AVFormatContext *    output_fmt_ctx= NULL;
AVStream *            audio_st = NULL;
AVCodec *            audio_codec = NULL;
double                audio_pts = 0.0;
AVFrame *            out_frame = avcodec_alloc_frame();

int                    audio_input_frame_size = 0;

uint8_t *            audio_data_buf = NULL;
uint8_t *            audio_out = NULL;
int                    audio_bit_rate;
int                    audio_sample_rate;
int                    audio_channels;

int decode_packet();
int open_audio_input(char* src_filename);
int decode_frame();

int open_encoder(char* output_filename);
AVStream *add_audio_stream(AVFormatContext *oc, AVCodec **codec,
    enum AVCodecID codec_id);
int open_audio(AVFormatContext *oc, AVCodec *codec, AVStream *st);
void close_audio(AVFormatContext *oc, AVStream *st);
void write_audio_frame(uint8_t ** audio_src_data, int audio_src_bufsize);

int open_audio_input(char* src_filename)
{
    int i =0;
    /* open input file, and allocate format context */
    if (avformat_open_input(&fmt_ctx, src_filename, NULL, NULL) < 0)
    {
        fprintf(stderr, "Could not open source file %s\n", src_filename);
        exit(1);
    }

    // Retrieve stream information
    if(avformat_find_stream_info(fmt_ctx, NULL)<0)
        return -1; // Couldn't find stream information

    // Dump information about file onto standard error
    av_dump_format(fmt_ctx, 0, src_filename, 0);

    // Find the first video stream
    for(i=0; i<fmt_ctx->nb_streams; i++)
    {
        if(fmt_ctx->streams[i]->codec->codec_type==AVMEDIA_TYPE_AUDIO)
        {
            audio_stream_index=i;
            break;
        }
    }
    if ( audio_stream_index != -1 )
    {
        // Get a pointer to the codec context for the audio stream
        codec_ctx_audio=fmt_ctx->streams[audio_stream_index]->codec;

        // Find the decoder for the video stream
        codec_audio=avcodec_find_decoder(codec_ctx_audio->codec_id);
        if(codec_audio==NULL) {
            fprintf(stderr, "Unsupported audio codec!\n");
            return -1; // Codec not found
        }

        // Open codec
        AVDictionary *codecDictOptions = NULL;
        if(avcodec_open2(codec_ctx_audio, codec_audio, &codecDictOptions)<0)
            return -1; // Could not open codec

        // Set up SWR context once you've got codec information
        swr = swr_alloc();
        av_opt_set_int(swr, "in_channel_layout",  codec_ctx_audio->channel_layout, 0);
        av_opt_set_int(swr, "out_channel_layout", codec_ctx_audio->channel_layout,  0);
        av_opt_set_int(swr, "in_sample_rate",     codec_ctx_audio->sample_rate, 0);
        av_opt_set_int(swr, "out_sample_rate",    codec_ctx_audio->sample_rate, 0);
        av_opt_set_sample_fmt(swr, "in_sample_fmt",  codec_ctx_audio->sample_fmt, 0);
        av_opt_set_sample_fmt(swr, "out_sample_fmt", AV_SAMPLE_FMT_S16,  0);
        swr_init(swr);

        // Allocate audio frame
        if ( decoded_frame == NULL ) decoded_frame = avcodec_alloc_frame();
        int nb_planes = 0;
        AVStream* audio_stream = fmt_ctx->streams[audio_stream_index];
        nb_planes = av_sample_fmt_is_planar(codec_ctx_audio->sample_fmt) ? codec_ctx_audio->channels : 1;
        int tempSize =  sizeof(uint8_t *) * nb_planes;
        audio_dst_data = (uint8_t**)av_mallocz(tempSize);
        if (!audio_dst_data)
        {
            fprintf(stderr, "Could not allocate audio data buffers\n");
        }
        else
        {
            for ( int i = 0 ; i < nb_planes ; i ++ )
            {
                audio_dst_data[i] = NULL;
            }
        }
    }
}


int decode_frame()
{
    int rv = 0;
    got_frame = 0;
    if ( fmt_ctx == NULL  )
    {
        return rv;
    }
    int ret = 0;
    audiobufsize = 0;
    rv = av_read_frame(fmt_ctx, &input_packet);
    if ( rv < 0 )
    {
        return rv;
    }
    rv = decode_packet();
    // Free the input_packet that was allocated by av_read_frame
    av_free_packet(&input_packet);
    return rv;
}

int decode_packet()
{
    int rv = 0;
    int ret = 0;

    //audio stream?
    if(input_packet.stream_index == audio_stream_index)
    {
        /* decode audio frame */
        rv = avcodec_decode_audio4(codec_ctx_audio, decoded_frame, &got_frame, &input_packet);
        if (rv < 0)
        {
            fprintf(stderr, "Error decoding audio frame\n");
            //return ret;
        }
        else
        {
            if (got_frame)
            {
                if ( audio_dst_data[0] == NULL )
                {
                     ret = av_samples_alloc(audio_dst_data, &audio_dst_linesize, decoded_frame->channels,
                        decoded_frame->nb_samples, (AVSampleFormat)decoded_frame->format, 1);
                    if (ret < 0)
                    {
                        fprintf(stderr, "Could not allocate audio buffer\n");
                        return AVERROR(ENOMEM);
                    }
                    /* TODO: extend return code of the av_samples_* functions so that this call is not needed */
                    audio_dst_bufsize = av_samples_get_buffer_size(NULL, audio_st->codec->channels,
                        decoded_frame->nb_samples, (AVSampleFormat)decoded_frame->format, 1);

                    //int16_t* outputBuffer = ...;
                    swr_convert(swr, audio_dst_data, out_frame->nb_samples,
                                (const uint8_t **)(decoded_frame->data), decoded_frame->nb_samples);
                    //swr_convert( swr, audio_dst_data, out_frame->nb_samples, (const uint8_t**) decoded_frame->extended_data, decoded_frame->nb_samples );
                }
                /* copy audio data to destination buffer:
                * this is required since rawaudio expects non aligned data */
                //av_samples_copy(audio_dst_data, decoded_frame->data, 0, 0,
                //    decoded_frame->nb_samples, decoded_frame->channels, (AVSampleFormat)decoded_frame->format);
            }
        }
    }
    return rv;
}


int open_encoder(char* output_filename )
{
    int rv = 0;

    /* allocate the output media context */
    AVOutputFormat *opfmt = NULL;

    avformat_alloc_output_context2(&output_fmt_ctx, opfmt, NULL, output_filename);
    if (!output_fmt_ctx) {
        printf("Could not deduce output format from file extension: using MPEG.\n");
        avformat_alloc_output_context2(&output_fmt_ctx, NULL, "mpeg", output_filename);
    }
    if (!output_fmt_ctx) {
        rv = -1;
    }
    else
    {
        output_format = output_fmt_ctx->oformat;
    }

    /* Add the audio stream using the default format codecs
    * and initialize the codecs. */
    audio_st = NULL;

    if ( output_fmt_ctx )
    {
        if (output_format->audio_codec != AV_CODEC_ID_NONE)
        {
            audio_st = add_audio_stream(output_fmt_ctx, &audio_codec, output_format->audio_codec);
        }

        /* Now that all the parameters are set, we can open the audio and
        * video codecs and allocate the necessary encode buffers. */
        if (audio_st)
        {
            rv = open_audio(output_fmt_ctx, audio_codec, audio_st);
            if ( rv < 0 ) return rv;
        }

        av_dump_format(output_fmt_ctx, 0, output_filename, 1);
        /* open the output file, if needed */
        if (!(output_format->flags & AVFMT_NOFILE))
        {
            if (avio_open(&output_fmt_ctx->pb, output_filename, AVIO_FLAG_WRITE) < 0) {
                fprintf(stderr, "Could not open '%s'\n", output_filename);
                rv = -1;
            }
            else
            {
                /* Write the stream header, if any. */
                if (avformat_write_header(output_fmt_ctx, NULL) < 0)
                {
                    fprintf(stderr, "Error occurred when opening output file\n");
                    rv = -1;
                }
            }
        }
    }

    return rv;
}

AVStream *add_audio_stream(AVFormatContext *oc, AVCodec **codec,
    enum AVCodecID codec_id)
{
    AVCodecContext *c;
    AVStream *st;

    /* find the audio encoder */
    *codec = avcodec_find_encoder(codec_id);
    if (!(*codec)) {
        fprintf(stderr, "Could not find codec\n");
        exit(1);
    }

    st = avformat_new_stream(oc, *codec);
    if (!st) {
        fprintf(stderr, "Could not allocate stream\n");
        exit(1);
    }
    st->id = 1;

    c = st->codec;

    /* put sample parameters */
    c->sample_fmt  = AV_SAMPLE_FMT_S16;
    c->bit_rate    = audio_bit_rate;
    c->sample_rate = audio_sample_rate;
    c->channels    = audio_channels;

    // some formats want stream headers to be separate
    if (oc->oformat->flags & AVFMT_GLOBALHEADER)
        c->flags |= CODEC_FLAG_GLOBAL_HEADER;

    return st;
}

int open_audio(AVFormatContext *oc, AVCodec *codec, AVStream *st)
{
    int ret=0;
    AVCodecContext *c;

    st->duration = fmt_ctx->duration;
    c = st->codec;

    /* open it */
    ret = avcodec_open2(c, codec, NULL) ;
    if ( ret < 0)
    {
        fprintf(stderr, "could not open codec\n");
        return -1;
        //exit(1);
    }

    if (c->codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE)
        audio_input_frame_size = 10000;
    else
        audio_input_frame_size = c->frame_size;
    int tempSize = audio_input_frame_size *
        av_get_bytes_per_sample(c->sample_fmt) *
        c->channels;
    return ret;
}

void close_audio(AVFormatContext *oc, AVStream *st)
{
    avcodec_close(st->codec);
}

void write_audio_frame(uint8_t ** audio_src_data, int audio_src_bufsize)
{
    AVFormatContext *oc = output_fmt_ctx;
    AVStream *st = audio_st;
    if ( oc == NULL || st == NULL ) return;
    AVCodecContext *c;
    AVPacket pkt = { 0 }; // data and size must be 0;
    int got_packet;

    av_init_packet(&pkt);
    c = st->codec;

    out_frame->nb_samples = audio_input_frame_size;
    int buf_size =         audio_src_bufsize *
        av_get_bytes_per_sample(c->sample_fmt) *
        c->channels;
    avcodec_fill_audio_frame(out_frame, c->channels, c->sample_fmt,
        (uint8_t *) *audio_src_data,
        buf_size, 1);
    avcodec_encode_audio2(c, &pkt, out_frame, &got_packet);
    if (!got_packet)
    {
    }
    else
    {
        if (pkt.pts != AV_NOPTS_VALUE)
            pkt.pts =  av_rescale_q(pkt.pts, st->codec->time_base, st->time_base);
        if (pkt.dts != AV_NOPTS_VALUE)
            pkt.dts = av_rescale_q(pkt.dts, st->codec->time_base, st->time_base);
        if ( c && c->coded_frame && c->coded_frame->key_frame)
            pkt.flags |= AV_PKT_FLAG_KEY;

         pkt.stream_index = st->index;
        pkt.flags |= AV_PKT_FLAG_KEY;
        /* Write the compressed frame to the media file. */
        if (av_interleaved_write_frame(oc, &pkt) != 0)
        {
            fprintf(stderr, "Error while writing audio frame\n");
            exit(1);
        }
    }
    av_free_packet(&pkt);
}


void write_delayed_frames(AVFormatContext *oc, AVStream *st)
{
    AVCodecContext *c = st->codec;
    int got_output = 0;
    int ret = 0;
    AVPacket pkt;
    pkt.data = NULL;
    pkt.size = 0;
    av_init_packet(&pkt);
    int i = 0;
    for (got_output = 1; got_output; i++)
    {
        ret = avcodec_encode_audio2(c, &pkt, NULL, &got_output);
        if (ret < 0)
        {
            fprintf(stderr, "error encoding frame\n");
            exit(1);
        }
        static int64_t tempPts = 0;
        static int64_t tempDts = 0;
        /* If size is zero, it means the image was buffered. */
        if (got_output)
        {
            if (pkt.pts != AV_NOPTS_VALUE)
                pkt.pts =  av_rescale_q(pkt.pts, st->codec->time_base, st->time_base);
            if (pkt.dts != AV_NOPTS_VALUE)
                pkt.dts = av_rescale_q(pkt.dts, st->codec->time_base, st->time_base);
            if ( c && c->coded_frame && c->coded_frame->key_frame)
                pkt.flags |= AV_PKT_FLAG_KEY;

            pkt.stream_index = st->index;
            /* Write the compressed frame to the media file. */
            ret = av_interleaved_write_frame(oc, &pkt);
        }
        else
        {
            ret = 0;
        }
        av_free_packet(&pkt);
    }
}

int main(int argc, char **argv)
{
    /* register all formats and codecs */
    av_register_all();
    avcodec_register_all();
    avformat_network_init();
    avdevice_register_all();
    int i =0;
    int ret=0;
    char src_filename[90] = "test_a.webm";
    char dst_filename[90] = "output.aac";
    open_audio_input(src_filename);
    if ( codec_ctx_audio->bit_rate == 0 ) codec_ctx_audio->bit_rate = 112000;
    audio_bit_rate        = codec_ctx_audio->bit_rate;
    audio_sample_rate    = codec_ctx_audio->sample_rate;
    audio_channels        = codec_ctx_audio->channels;
    open_encoder( dst_filename );
    while(1)
    {
        int rv = decode_frame();
        if ( rv < 0 )
        {
            break;
        }

        if (audio_st)
        {
            audio_pts = (double)audio_st->pts.val * audio_st->time_base.num /
                audio_st->time_base.den;
        }
        else
        {
            audio_pts = 0.0;
        }
        if ( codec_ctx_audio )
        {
            if ( got_frame)
            {
                write_audio_frame( audio_dst_data, audio_dst_bufsize );
            }
        }
        if ( audio_dst_data[0] )
        {
            av_freep(&audio_dst_data[0]);
            audio_dst_data[0] = NULL;
        }
        printf("\naudio_pts: %.3f", audio_pts);
    }
    while(1)
    {
        if ( audio_dst_data && audio_dst_data[0] )
        {
            av_freep(&audio_dst_data[0]);
            audio_dst_data[0] = NULL;
        }
        ret = av_samples_alloc(audio_dst_data, NULL, codec_ctx_audio->channels,
            decoded_frame->nb_samples, AV_SAMPLE_FMT_S16, 0);
        ret = swr_convert(swr, audio_dst_data, out_frame->nb_samples,NULL, 0);
        if ( ret <= 0 ) break;
        write_audio_frame( audio_dst_data, audio_dst_bufsize );
    }
    write_delayed_frames( output_fmt_ctx, audio_st );
    av_write_trailer(output_fmt_ctx);
    close_audio( output_fmt_ctx, audio_st);
    swr_free(&swr);
    avcodec_free_frame(&out_frame);
    getch();
    return 0;
}

"test_a.webm" input file results in longer duration (40 second output), and if I change it to "jet.webm", it is converted fine.

Both input files are approximately 18 second duration.

For reference, these files can be downloaded from links below:

http://www.filedropper.com/testa , http://www.filedropper.com/jet

Alternatively, they are zipped and uploaded elsewhere as well:

http://www.files.com/shared/52c3eefe990ea/test_audio_files.zip

Could someone kindly guide on what I am doing wrong here?

Thanks in advance...

p.s. These files are taken/extracted from different online sources/demos.

Edit 2-1-14: After debugging, I can see audio_pts variable is being populated incorrectly. It relies on audio_st->pts.val, which is automatically calculated upon calling av_interleaved_write_frame() function. I cannot step inside av_interleaved_write_frame() function since I am on Windows, using libav dlls/libs.

So,

For jet.webm file (whose conversion is fine), audio_st->pts.val goes till maximum: 1665567, and audio_pts becomes:

1665567*1/90000 = 18.5063

For test_a.webm file (whose conversion is bad), audio_st->pts.val goes till maximum: 3606988, and audio_pts becomes:

3606988*1/90000 = 40.0776

• reference line: audio_pts = (double)audio_st->pts.val * audio_st->time_base.num / audio_st->time_base.den;

Since PTS is very off, it shouldn't be playing fine logically as well. But I cannot say av_interleaved_write_frame() function is doing it wrong; surely something cleaner can be managed on my end.

Edit 3-1-14: Discovered one more thing: while reading jet.webm file, decoded frame's nb_sample are always 1024 (except for 1st frame: 576), but in case of test_a.webm file, nb_samples are either 1024, or 128, with exceptions of 576 (less frequent). If I ignore writing of frame when nb_sample is 128, I get approximately same file length in the end, but you can hear bits and pieces are missing here and there.

How can I deal with this?