FFmpeg encoding aac audio, encoded file can not be played

Question

I am trying to encodea aac audio. Example of MP2 here . I followed this documentation. I encode the audio with the code below by calling startEncoding and after 2 seconds I call stopEncoding . Everything seems to work fine I get an file with some size but the problem is I can not open or play it. I dont know why. I must be doing something wrong in the code.

header:

class MediaEncoder {
public:
    MediaEncoder(char *filePath);

    void startEncoding();
    void stopEncoding();

    void encode(AVFrame *frame);
    bool isEncoding() const;

    void startEncoderWorker();

    int32_t check_sample_fmt(enum AVSampleFormat sample_fmt);
    
    bool signalExitFuture = false;
    int32_t ret;

private:
    std::future<void> encodingFuture;
    AVCodecContext *avCodecContext;
    AVFrame *avFrame;
    AVPacket *avPacket;
    AVCodec *codec;
    FILE* file;
};

cpp:

MediaEncoder::MediaEncoder(char *filePath){
    buffer = std::make_unique<LockFreeQueue<float>>(recorderBufferSize);

    /* find the encoder */
    codec = avcodec_find_encoder(AV_CODEC_ID_AAC);
    avCodecContext = avcodec_alloc_context3(codec);

    avCodecContext->bit_rate = 64000;

    /* check that the encoder supports given sample format */
    avCodecContext->sample_fmt = AVSampleFormat::AV_SAMPLE_FMT_FLTP;
    
    /* select other audio parameters supported by the encoder */
    avCodecContext->sample_rate = defaultSampleRate;
    avCodecContext->channel_layout = AV_CH_LAYOUT_STEREO;
    avCodecContext->channels = av_get_channel_layout_nb_channels(avCodecContext->channel_layout);

    /* open it */
    avcodec_open2(avCodecContext, codec, nullptr)
    
    file = fopen(filePath, "wb");
    
    /* packet for holding encoded output */
    avPacket = av_packet_alloc();
    
    /* frame containing input raw audio */
    avFrame = av_frame_alloc();
    
    avFrame->nb_samples = avCodecContext->frame_size;
    avFrame->format = avCodecContext->sample_fmt;
    avFrame->channel_layout = avCodecContext->channel_layout;

    /* allocate the data buffers */
    av_frame_get_buffer(avFrame, 0);
}


void MediaEncoder::startEncoding() {
    // set flags for decoding thread
    signalExitFuture = false;

    encodingFuture = std::async(std::launch::async, &MediaEncoder::startEncoderWorker, this);
}

void MediaEncoder::stopEncoding() {
    signalExitFuture = true;
}

bool MediaEncoder::isEncoding() const {
    return !signalExitFuture;
}

void MediaEncoder::encode(AVFrame *frame) {
    /* send the frame for encoding */
    ret = avcodec_send_frame(avCodecContext, frame);
    if (ret < 0) {
        LOGE("Error sending the frame to the encoder %s",
             av_err2str(ret));
        *recorderStatePointer = MediaEncoderState::RUNTIME_FAIL;
        return;
    }

    /* read all the available output packets (in general there may be any
     * number of them */
    while (ret >= 0) {
        ret = avcodec_receive_packet(avCodecContext, avPacket);
        if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) {
            return;
        } else if (ret < 0) {
            LOGE("Error encoding audio frame %s",
                 av_err2str(ret));
            *recorderStatePointer = MediaEncoderState::RUNTIME_FAIL;
            return;
        }
       
        /* Solution begins here
        int aac_profile = 2;           // AAC LC
        int frequencey_index = 4;      // 44,100Hz
        int channel_configuration = 2; // stereo (left, right)

        int frame_length = avPacket->size + 7;              // your frame length
     
        unsigned char adts_header[7];

        // fill in ADTS data
        adts_header[0] = (unsigned char) 0xFF;
        adts_header[1] = (unsigned char) 0xF9;
        adts_header[2] = (unsigned char) (((aac_profile -1) << 6 ) + (frequencey_index << 2) + (channel_configuration >> 2));
        adts_header[3] = (unsigned char) (((channel_configuration & 3) << 6) + (frame_length >> 11));
        adts_header[4] = (unsigned char) ((frame_length & 0x7FF) >> 3);
        adts_header[5] = (unsigned char) (((frame_length & 7) << 5) + 0x1F);
        adts_header[6] = (unsigned char) 0xFC;

        fwrite(adts_header, 1, 7, file); 
        Solution ends here */ 
        fwrite(avPacket->data, 1, avPacket->size, file);
        av_packet_unref(avPacket);
    }
}

void MediaEncoder::startEncoderWorker() {
    try {
        float *leftChannel;
        float *rightChannel;
        float val;

        while (!signalExitFuture) {
            ret = av_frame_make_writable(avFrame);
            if (ret < 0) {
                LOGE("av_frame_make_writable Can not Ensure that the frame data is writable. %s",
                     av_err2str(ret));
                *recorderStatePointer = MediaEncoderState::RUNTIME_FAIL;
                return;
            }

            leftChannel = (float *) avFrame->data[0];
            rightChannel = (float *) avFrame->data[1];

            for (int32_t i = 0; i < avCodecContext->frame_size; ++i) {
               
                leftChannel[i] = 0.4;
                rightChannel[i] = 0.4;
            }

            encode(avFrame);
        }

        /* flush the encoder */
        encode(nullptr);

        fclose(file);
        LOGE("Encoding finished!");

        av_frame_free(&avFrame);
        av_packet_free(&avPacket);
        avcodec_free_context(&avCodecContext);
    } catch (std::exception &e) {
        LOGE("startEncoderWorker uncaught exception %s", e.what());
    }

    LOGE("Deleting Media Encoder!");

}

Here is an recorded 11 seconds of an aac file recorded with real float pcm data rather than 0.4 as it is in the code I posted. Google Drive Link

The code above works. Thanks to the legend @Markus-Schumann

Answer 1

I assume that the AAC encoder outputs raw AAC frames. Writing the raw AAC frames to file will not create a playable output. So you have to add a frame headers or mux your output into a MP4 file. Look for ADIF, ADTS or MP4 as file format.

Let's assume you'd like to do ADTS:

int aac_profile = 2;           // AAC LC
int frequencey_index = 4;      // 44,100Hz
int channel_configuration = 2; // stereo (left, right)

// https://wiki.multimedia.cx/index.php/MPEG-4_Audio#Channel_Configurations

int frame_length;              // your frame length + 7 bytes for the header


unsigned char adts_header[7];
// Take look here: https://wiki.multimedia.cx/index.php/ADTS

// fill in ADTS data
adts_header[0] = (unsigned char) 0xFF;
adts_header[1] = (unsigned char) 0xF9;
adts_header[2] = (unsigned char) (((aac_profile -1) << 6 ) + (frequencey_index << 2) + (channel_configuration >> 2));
adts_header[3] = (unsigned char) (((channel_configuration & 3) << 6) + (frame_length >> 11));
adts_header[4] = (unsigned char) ((frame_length & 0x7FF) >> 3);
adts_header[5] = (unsigned char) (((frame_length & 7) << 5) + 0x1F);
adts_header[6] = (unsigned char) 0xFC;

So you have to prefix each frame that comes out of the encoder with the above header make sure to set frame_length to the AAC buffer length coming out of the encoder. On disk it should look like: |ADTS header|AAC frame|ADTS header|AAC frame|...|ADTS header|AAC frame|