video-pc/libs/Recorder/muxer_ffmpeg.cpp

#include "muxer_ffmpeg.h"
#include "headers_ffmpeg.h"

#include "muxer_define.h"

#include "encoder_video.h"
#include "encoder_video_factory.h"
#include "record_desktop.h"
#include "sws_helper.h"

#include "encoder_aac.h"
#include "filter_amix.h"
#include "filter_aresample.h"
#include "record_audio.h"

#include "ring_buffer.h"

#include "error_define.h"
#include "log_helper.h"

namespace am {
muxer_ffmpeg::muxer_ffmpeg()
{
    ffmpeg_register_all();
    ffmpeg_register_devices(); // 添加设备注册，支持RTSP等网络协议

    _v_stream = NULL;
    _a_stream = NULL;

    _fmt = NULL;
    _fmt_ctx = NULL;

    _base_time = -1;
}

muxer_ffmpeg::~muxer_ffmpeg()
{
    stop();
    cleanup();
}

int muxer_ffmpeg::init(const char *output_file,
                       record_desktop *source_desktop,
                       record_audio **source_audios,
                       const int source_audios_nb,
                       const MUX_SETTING_T &setting)
{
    int error = AE_NO;
    int ret = 0;

    do {
        al_info("start to initialize muxer ,output:%s ", output_file);

        error = alloc_oc(output_file, setting);
        if (error != AE_NO)
            break;

        if (_fmt->video_codec != AV_CODEC_ID_NONE) {
            error = add_video_stream(setting, source_desktop);
            if (error != AE_NO)
                break;
        }

        if (_fmt->audio_codec != AV_CODEC_ID_NONE && source_audios_nb) {
            error = add_audio_stream(setting, source_audios, source_audios_nb);
            if (error != AE_NO)
                break;
        }

        error = open_output(output_file, setting);
        if (error != AE_NO)
            break;

        av_dump_format(_fmt_ctx, 0, NULL, 1);

        _inited = true;

    } while (0);

    if (error != AE_NO) {
        cleanup();
        al_debug("muxer ffmpeg initialize failed:%s %d", err2str(error), ret);
    }

    return error;
}

int muxer_ffmpeg::start()
{
    std::lock_guard<std::mutex> lock(_mutex);

    int error = AE_NO;

    if (_running == true) {
        return AE_NO;
    }

    if (_inited == false) {
        return AE_NEED_INIT;
    }

    _base_time = av_gettime_relative();

    if (_v_stream && _v_stream->v_enc)
        _v_stream->v_enc->start();

    if (_a_stream && _a_stream->a_enc)
        _a_stream->a_enc->start();

    if (_a_stream && _a_stream->a_nb >= 2 && _a_stream->a_filter_amix)
        _a_stream->a_filter_amix->start();

    if (_a_stream && _a_stream->a_nb < 2 && _a_stream->a_filter_aresample) {
        for (int i = 0; i < _a_stream->a_nb; i++) {
            _a_stream->a_filter_aresample[i]->start();
        }
    }

    if (_a_stream && _a_stream->a_src) {
        for (int i = 0; i < _a_stream->a_nb; i++) {
            if (_a_stream->a_src[i])
                _a_stream->a_src[i]->start();
        }
    }

    if (_v_stream && _v_stream->v_src)
        _v_stream->v_src->start();

    _running = true;

    return error;
}

int muxer_ffmpeg::stop()
{
    std::lock_guard<std::mutex> lock(_mutex);

    if (_running == false)
        return AE_NO;

    _running = false;

    al_debug("try to stop muxer....");

    al_debug("stop audio recorder...");
    if (_a_stream && _a_stream->a_src) {
        for (int i = 0; i < _a_stream->a_nb; i++) {
            _a_stream->a_src[i]->stop();
        }
    }

    al_debug("stop video recorder...");
    if (_v_stream && _v_stream->v_src)
        _v_stream->v_src->stop();

    al_debug("stop audio amix filter...");
    if (_a_stream && _a_stream->a_filter_amix)
        _a_stream->a_filter_amix->stop();

    al_debug("stop audio aresampler filter...");
    if (_a_stream && _a_stream->a_filter_aresample) {
        for (int i = 0; i < _a_stream->a_nb; i++) {
            _a_stream->a_filter_aresample[i]->stop();
        }
    }

    al_debug("stop video encoder...");
    if (_v_stream && _v_stream->v_enc)
        _v_stream->v_enc->stop();

    al_debug("stop audio encoder...");
    if (_a_stream) {
        if (_a_stream->a_enc)
            _a_stream->a_enc->stop();
    }

    al_debug("write file trailer...");
    if (_fmt_ctx)
        av_write_trailer(_fmt_ctx); //must write trailer ,otherwise file can not play

    al_debug("muxer stopped...");

    return AE_NO;
}

int muxer_ffmpeg::pause()
{
    _paused = true;

    return 0;
}

int muxer_ffmpeg::resume()
{
    _paused = false;
    return 0;
}

// Exposed getters for encoder instances
encoder_video* muxer_ffmpeg::get_video_encoder() const
{
    return _v_stream ? _v_stream->v_enc : nullptr;
}

encoder_aac* muxer_ffmpeg::get_audio_encoder() const
{
    return _a_stream ? _a_stream->a_enc : nullptr;
}

void muxer_ffmpeg::on_desktop_data(AVFrame *frame)
{
    if (_running == false || _paused == true || !_v_stream || !_v_stream->v_enc
        || !_v_stream->v_sws) {
        return;
    }

    int len = 0, ret = AE_NO;
    uint8_t *yuv_data = NULL;

    ret = _v_stream->v_sws->convert(frame, &yuv_data, &len);

    if (ret == AE_NO && yuv_data && len) {
        _v_stream->v_enc->put(yuv_data, len, frame);

        if (_on_yuv_data && _preview_enabled == true)
            _on_yuv_data(yuv_data, len, frame->width, frame->height, 0);
    }
}

void muxer_ffmpeg::on_desktop_error(int error)
{
    al_fatal("on desktop capture error:%d", error);
}

int getPcmDB(const unsigned char *pcmdata, size_t size)
{
    int db = 0;
    float value = 0;
    double sum = 0;
    double average = 0;
    int bit_per_sample = 32;
    int byte_per_sample = bit_per_sample / 8;
    int channel_num = 2;

    for (int i = 0; i < size; i += channel_num * byte_per_sample) {
        memcpy(&value, pcmdata + i, byte_per_sample);
        sum += abs(value);
    }
    average = sum / (double) (size / byte_per_sample / channel_num);
    if (average > 0) {
        db = (int) (20 * log10f(average));
    }

    al_debug("%d   %f     %f", db, average, sum);
    return db;
}

static int pcm_fltp_db_count(AVFrame *frame, int channels)
{
    int i = 0, ch = 0;
    int ndb = 0;
    float value = 0.;
    float *ch_left = (float *) frame->data[0];
    //float *ch_right = (float *)frame->data[1];
    for (i = 0; i < frame->nb_samples; i++) {
        value += fabs(ch_left[i]);
    }

    value = value / frame->nb_samples;
    if (0 != value) {
        ndb = (int) (20.0 * log10((value / 1.0)));
    } else
        ndb = -100;

    return ndb;
}

void muxer_ffmpeg::on_audio_data(AVFrame *frame, int index)
{
    if (_running == false || _paused == true)
        return;

    if (_a_stream->a_filter_amix != nullptr)
        _a_stream->a_filter_amix->add_frame(frame, index);
    else if (_a_stream->a_filter_aresample != nullptr
             && _a_stream->a_filter_aresample[index] != nullptr) {
        _a_stream->a_filter_aresample[index]->add_frame(frame);
    }

    return;
}

void muxer_ffmpeg::on_audio_error(int error, int index)
{
    al_fatal("on audio capture error:%d with stream index:%d", error, index);
}

void muxer_ffmpeg::on_filter_amix_data(AVFrame *frame, int)
{
    if (_running == false || !_a_stream->a_enc)
        return;

    AUDIO_SAMPLE *resamples = _a_stream->a_resamples[0];

    int copied_len = 0;
    int sample_len = ffmpeg_get_buffer_size((AVSampleFormat) frame->format,
                                            ffmpeg_get_frame_channels(frame),
                                            frame->nb_samples,
                                            1);
    sample_len = ffmpeg_get_buffer_size((AVSampleFormat) frame->format,
                                        ffmpeg_get_frame_channels(frame),
                                        frame->nb_samples,
                                        1);

#ifdef _DEBUG
    //al_debug("dg:%d", pcm_fltp_db_count(frame, frame->channels));
#endif

    int remain_len = sample_len;

    //for data is planar,should copy data[0] data[1] to correct buff pos
    if (av_sample_fmt_is_planar((AVSampleFormat) frame->format) == 0) {
        while (remain_len > 0) {
            //cache pcm
            copied_len = min(resamples->size - resamples->sample_in, remain_len);
            if (copied_len) {
                memcpy(resamples->buff + resamples->sample_in,
                       frame->data[0] + sample_len - remain_len,
                       copied_len);
                resamples->sample_in += copied_len;
                remain_len = remain_len - copied_len;
            }

            //got enough pcm to encoder,resample and mix
            if (resamples->sample_in == resamples->size) {
                _a_stream->a_enc->put(resamples->buff, resamples->size, frame);

                resamples->sample_in = 0;
            }
        }
    } else { // planar: copy each channel plane into contiguous planes in buffer
        const int channels = ffmpeg_get_frame_channels(frame);
        while (remain_len > 0) {
            copied_len = min(resamples->size - resamples->sample_in, remain_len);
            if (copied_len) {
                const int copied_per_plane = copied_len / channels;
                const int written_per_plane = resamples->sample_in / channels;
                const int plane_size = resamples->size / channels;
                const int src_offset = (sample_len - remain_len) / channels;
                for (int ch = 0; ch < channels; ++ch) {
                    memcpy(resamples->buff + ch * plane_size + written_per_plane,
                           frame->data[ch] + src_offset,
                           copied_per_plane);
                }
                resamples->sample_in += copied_len;
                remain_len = remain_len - copied_len;
            }

            //got enough pcm to encoder,resample and mix
            if (resamples->sample_in == resamples->size) {
                _a_stream->a_enc->put(resamples->buff, resamples->size, frame);

                resamples->sample_in = 0;
            }
        }
    }
}

void muxer_ffmpeg::on_filter_amix_error(int error, int)
{
    al_fatal("on filter amix audio error:%d", error);
}

void muxer_ffmpeg::on_filter_aresample_data(AVFrame *frame, int index)
{
    if (_running == false || !_a_stream->a_enc)
        return;

    AUDIO_SAMPLE *resamples = _a_stream->a_resamples[index];

    int copied_len = 0;
    int sample_len = ffmpeg_get_buffer_size((AVSampleFormat) frame->format,
                                            ffmpeg_get_frame_channels(frame),
                                            frame->nb_samples,
                                            1);
    sample_len = ffmpeg_get_buffer_size((AVSampleFormat) frame->format,
                                        ffmpeg_get_frame_channels(frame),
                                        frame->nb_samples,
                                        1);

    int remain_len = sample_len;

    //for data is planar,should copy data[0] data[1] to correct buff pos
    if (av_sample_fmt_is_planar((AVSampleFormat) frame->format) == 0) {
        while (remain_len > 0) {
            //cache pcm
            copied_len = min(resamples->size - resamples->sample_in, remain_len);
            if (copied_len) {
                memcpy(resamples->buff + resamples->sample_in,
                       frame->data[0] + sample_len - remain_len,
                       copied_len);
                resamples->sample_in += copied_len;
                remain_len = remain_len - copied_len;
            }

            //got enough pcm to encoder,resample and mix
            if (resamples->sample_in == resamples->size) {
                _a_stream->a_enc->put(resamples->buff, resamples->size, frame);

                resamples->sample_in = 0;
            }
        }
    } else { // planar: copy each channel plane into contiguous planes in buffer
        const int channels = ffmpeg_get_frame_channels(frame);
        while (remain_len > 0) {
            copied_len = min(resamples->size - resamples->sample_in, remain_len);

            if (copied_len) {
                const int copied_per_plane = copied_len / channels;
                const int written_per_plane = resamples->sample_in / channels;
                const int plane_size = resamples->size / channels;
                const int src_offset = (sample_len - remain_len) / channels;
                for (int ch = 0; ch < channels; ++ch) {
                    memcpy(resamples->buff + ch * plane_size + written_per_plane,
                           frame->data[ch] + src_offset,
                           copied_per_plane);
                }
                resamples->sample_in += copied_len;
                remain_len = remain_len - copied_len;
            }

            if (resamples->sample_in == resamples->size) {
                _a_stream->a_enc->put(resamples->buff, resamples->size, frame);

                resamples->sample_in = 0;
            }
        }
    }
}

void muxer_ffmpeg::on_filter_aresample_error(int error, int index)
{
    al_fatal("on filter aresample[%d] audio error:%d", index, error);
}

void muxer_ffmpeg::on_enc_264_data(AVPacket *packet)
{
    if (_running && _v_stream) {
        write_video(packet);
    }
}

void muxer_ffmpeg::on_enc_264_error(int error)
{
    al_fatal("on desktop encode error:%d", error);
}

void muxer_ffmpeg::on_enc_aac_data(AVPacket *packet)
{
    if (_running && _a_stream) {
        write_audio(packet);
    }
}

void muxer_ffmpeg::on_enc_aac_error(int error)
{
    al_fatal("on audio encode error:%d", error);
}

int muxer_ffmpeg::alloc_oc(const char *output_file, const MUX_SETTING_T &setting)
{
    _output_file = std::string(output_file);

    int error = AE_NO;
    int ret = 0;

    do {
        // 检查协议类型并指定相应的输出格式
        const char* format_name = NULL;
        std::string url_str(output_file);
        if (url_str.find("rtmp://") == 0 || url_str.find("rtmps://") == 0) {
            format_name = "flv";
        } else if (url_str.find("rtsp://") == 0) {
            // RTSP推流使用RTSP muxer，由FFmpeg管理RTP会话
            format_name = "rtsp";
            al_debug("RTSP URL detected, using RTSP format");
        }
        
        ret = avformat_alloc_output_context2(&_fmt_ctx, NULL, format_name, output_file);
        if (ret < 0 || !_fmt_ctx) {
            al_debug("avformat_alloc_output_context2 failed with ret=%d, format=%s, url=%s", 
                     ret, format_name ? format_name : "auto", output_file);
            error = AE_FFMPEG_ALLOC_CONTEXT_FAILED;
            break;
        }

        _fmt = _fmt_ctx->oformat;
    } while (0);

    return error;
}

int muxer_ffmpeg::add_video_stream(const MUX_SETTING_T &setting, record_desktop *source_desktop)
{
    int error = AE_NO;
    int ret = 0;

    _v_stream = new MUX_STREAM();
    memset(_v_stream, 0, sizeof(MUX_STREAM));

    _v_stream->v_src = source_desktop;

    _v_stream->pre_pts = -1;

    _v_stream->v_src->registe_cb(std::bind(&muxer_ffmpeg::on_desktop_data,
                                           this,
                                           std::placeholders::_1),
                                 std::bind(&muxer_ffmpeg::on_desktop_error,
                                           this,
                                           std::placeholders::_1));

    RECORD_DESKTOP_RECT v_rect = _v_stream->v_src->get_rect();

    do {
        error = encoder_video_new(setting.v_encoder_id, &_v_stream->v_enc);
        if (error != AE_NO)
            break;

        error = _v_stream->v_enc->init(setting.v_out_width,
                                       setting.v_out_height,
                                       setting.v_frame_rate,
                                       setting.v_bit_rate,
                                       setting.v_qb);
        if (error != AE_NO)
            break;

        _v_stream->v_enc->registe_cb(std::bind(&muxer_ffmpeg::on_enc_264_data,
                                               this,
                                               std::placeholders::_1),
                                     std::bind(&muxer_ffmpeg::on_enc_264_error,
                                               this,
                                               std::placeholders::_1));

        _v_stream->v_sws = new sws_helper();
        error = _v_stream->v_sws->init(_v_stream->v_src->get_pixel_fmt(),
                                       v_rect.right - v_rect.left,
                                       v_rect.bottom - v_rect.top,
                                       AV_PIX_FMT_YUV420P,
                                       setting.v_out_width,
                                       setting.v_out_height);
        if (error != AE_NO)
            break;

        const AVCodec *codec = avcodec_find_encoder(_v_stream->v_enc->get_codec_id());
        if (!codec) {
            error = AE_FFMPEG_FIND_ENCODER_FAILED;
            break;
        }

        // FFmpeg 7兼容性：不再直接修改AVOutputFormat的video_codec字段
        // 编码器信息通过AVStream的codecpar设置

        AVStream *st = avformat_new_stream(_fmt_ctx, codec);
        if (!st) {
            error = AE_FFMPEG_NEW_STREAM_FAILED;
            break;
        }

        ffmpeg_set_stream_codec_id(st, _v_stream->v_enc->get_codec_id());
        ffmpeg_set_stream_bit_rate(st, setting.v_bit_rate);
        ffmpeg_set_stream_codec_type(st, AVMEDIA_TYPE_VIDEO);

        // 使用编码器/设置的帧率作为时间基（1/fps），避免异常的帧率显示
        st->time_base = {1, setting.v_frame_rate};
        ffmpeg_set_stream_pix_fmt(st, AV_PIX_FMT_YUV420P);

        ffmpeg_set_stream_dimensions(st, setting.v_out_width, setting.v_out_height);

        // 正确设置平均帧率为 fps/1
        st->avg_frame_rate = {setting.v_frame_rate, 1};

        // 始终为视频流设置extradata（SPS/PPS等），RTSP/SDP需要该信息
        {
            uint8_t *extradata = (uint8_t *) av_memdup(_v_stream->v_enc->get_extradata(),
                                                       _v_stream->v_enc->get_extradata_size());
            ffmpeg_set_stream_extradata(st, extradata, _v_stream->v_enc->get_extradata_size());
        }

         _v_stream->st = st;

         _v_stream->setting = setting;
         //_v_stream->filter = av_bitstream_filter_init("h264_mp4toannexb");
    } while (0);

    return error;
}

int muxer_ffmpeg::add_audio_stream(const MUX_SETTING_T &setting,
                                   record_audio **source_audios,
                                   const int source_audios_nb)
{
    int error = AE_NO;
    int ret = 0;

    _a_stream = new MUX_STREAM();
    memset(_a_stream, 0, sizeof(MUX_STREAM));

    _a_stream->a_nb = source_audios_nb;
    _a_stream->a_filter_aresample = new filter_aresample *[_a_stream->a_nb];
    _a_stream->a_resamples = new AUDIO_SAMPLE *[_a_stream->a_nb];
    _a_stream->a_samples = new AUDIO_SAMPLE *[_a_stream->a_nb];
    _a_stream->a_src = new record_audio *[_a_stream->a_nb];
    _a_stream->pre_pts = -1;

    do {
        _a_stream->a_enc = new encoder_aac();
        error = _a_stream->a_enc->init(setting.a_nb_channel,
                                       setting.a_sample_rate,
                                       setting.a_sample_fmt,
                                       setting.a_bit_rate);
        if (error != AE_NO)
            break;

        _a_stream->a_enc->registe_cb(std::bind(&muxer_ffmpeg::on_enc_aac_data,
                                               this,
                                               std::placeholders::_1),
                                     std::bind(&muxer_ffmpeg::on_enc_aac_error,
                                               this,
                                               std::placeholders::_1));

        for (int i = 0; i < _a_stream->a_nb; i++) {
            _a_stream->a_src[i] = source_audios[i];
            _a_stream->a_src[i]->registe_cb(std::bind(&muxer_ffmpeg::on_audio_data,
                                                      this,
                                                      std::placeholders::_1,
                                                      std::placeholders::_2),
                                            std::bind(&muxer_ffmpeg::on_audio_error,
                                                      this,
                                                      std::placeholders::_1,
                                                      std::placeholders::_2),
                                            i);

            _a_stream->a_filter_aresample[i] = new filter_aresample();
            _a_stream->a_resamples[i] = new AUDIO_SAMPLE({NULL, 0, 0});

            FILTER_CTX ctx_in = {0}, ctx_out = {0};
            ctx_in.time_base = _a_stream->a_src[i]->get_time_base();
            ctx_in.channel_layout = ffmpeg_get_default_channel_layout(
                _a_stream->a_src[i]->get_channel_num());
            ctx_in.nb_channel = _a_stream->a_src[i]->get_channel_num();
            ctx_in.sample_fmt = _a_stream->a_src[i]->get_fmt();
            ctx_in.sample_rate = _a_stream->a_src[i]->get_sample_rate();

            ctx_out.time_base = {1, AV_TIME_BASE};
            ctx_out.channel_layout = ffmpeg_get_default_channel_layout(setting.a_nb_channel);
            ctx_out.nb_channel = setting.a_nb_channel;
            ctx_out.sample_fmt = setting.a_sample_fmt;
            ctx_out.sample_rate = setting.a_sample_rate;

            _a_stream->a_filter_aresample[i]->init(ctx_in, ctx_out, i);

            _a_stream->a_filter_aresample[i]->registe_cb(
                std::bind(&muxer_ffmpeg::on_filter_aresample_data,
                          this,
                          std::placeholders::_1,
                          std::placeholders::_2),
                std::bind(&muxer_ffmpeg::on_filter_aresample_error,
                          this,
                          std::placeholders::_1,
                          std::placeholders::_2));

            _a_stream->a_resamples[i]->size
                = av_samples_get_buffer_size(NULL,
                                             setting.a_nb_channel,
                                             _a_stream->a_enc->get_nb_samples(),
                                             setting.a_sample_fmt,
                                             1);
            _a_stream->a_resamples[i]->buff = new uint8_t[_a_stream->a_resamples[i]->size];
            _a_stream->a_samples[i] = new AUDIO_SAMPLE({NULL, 0, 0});
            _a_stream->a_samples[i]->size
                = av_samples_get_buffer_size(NULL,
                                             _a_stream->a_src[i]->get_channel_num(),
                                             _a_stream->a_enc->get_nb_samples(),
                                             _a_stream->a_src[i]->get_fmt(),
                                             1);
            _a_stream->a_samples[i]->buff = new uint8_t[_a_stream->a_samples[i]->size];
        }

        if (_a_stream->a_nb >= 2) {
            _a_stream->a_filter_amix = new am::filter_amix();
            error = _a_stream->a_filter_amix->init({NULL,
                                                    NULL,
                                                    _a_stream->a_src[0]->get_time_base(),
                                                    _a_stream->a_src[0]->get_sample_rate(),
                                                    _a_stream->a_src[0]->get_fmt(),
                                                    _a_stream->a_src[0]->get_channel_num(),
                                                    (int64_t)ffmpeg_get_default_channel_layout(
                                                        _a_stream->a_src[0]->get_channel_num())},
                                                   {NULL,
                                                    NULL,
                                                    _a_stream->a_src[1]->get_time_base(),
                                                    _a_stream->a_src[1]->get_sample_rate(),
                                                    _a_stream->a_src[1]->get_fmt(),
                                                    _a_stream->a_src[1]->get_channel_num(),
                                                    (int64_t)ffmpeg_get_default_channel_layout(
                                                        _a_stream->a_src[1]->get_channel_num())},
                                                   {NULL,
                                                    NULL,
                                                    {1, AV_TIME_BASE},
                                                    setting.a_sample_rate,
                                                    setting.a_sample_fmt,
                                                    setting.a_nb_channel,
                                                    (int64_t)ffmpeg_get_default_channel_layout(
                                                        setting.a_nb_channel)});

            if (error != AE_NO) {
                break;
            }

            _a_stream->a_filter_amix->registe_cb(std::bind(&muxer_ffmpeg::on_filter_amix_data,
                                                           this,
                                                           std::placeholders::_1,
                                                           std::placeholders::_2),
                                                 std::bind(&muxer_ffmpeg::on_filter_amix_error,
                                                           this,
                                                           std::placeholders::_1,
                                                           std::placeholders::_2));
        }

        const AVCodec *codec = avcodec_find_encoder(_a_stream->a_enc->get_codec_id());
        if (!codec) {
            error = AE_FFMPEG_FIND_ENCODER_FAILED;
            break;
        }

        // FFmpeg 7兼容性：不再直接修改AVOutputFormat的codec字段
        // 这些字段在新版本中是只读的，编码器信息通过AVStream设置
        AVCodecID audio_codec_id = _a_stream->a_enc->get_codec_id();

        AVStream *st = avformat_new_stream(_fmt_ctx, codec);
        if (!st) {
            error = AE_FFMPEG_NEW_STREAM_FAILED;
            break;
        }

        av_dict_set(&st->metadata, "title", "Track1", 0);

        // 设置音频流编码器参数
        ffmpeg_set_stream_codec_id(st, audio_codec_id);
        ffmpeg_set_stream_codec_type(st, AVMEDIA_TYPE_AUDIO);
        ffmpeg_set_stream_bit_rate(st, setting.a_bit_rate);
        
        // 设置采样率到AVStream的codecpar中（FFmpeg 7兼容性）
#if FFMPEG_VERSION_MAJOR >= 7
        st->codecpar->sample_rate = setting.a_sample_rate;
        av_channel_layout_default(&st->codecpar->ch_layout, setting.a_nb_channel);
        st->codecpar->format = setting.a_sample_fmt;
#elif FFMPEG_VERSION_MAJOR >= 4
        st->codecpar->sample_rate = setting.a_sample_rate;
        st->codecpar->channels = setting.a_nb_channel;
        st->codecpar->channel_layout = ffmpeg_get_default_channel_layout(setting.a_nb_channel);
        st->codecpar->format = setting.a_sample_fmt;
#else
        st->codec->sample_rate = setting.a_sample_rate;
        st->codec->channels = setting.a_nb_channel;
        st->codec->channel_layout = ffmpeg_get_default_channel_layout(setting.a_nb_channel);
        st->codec->sample_fmt = setting.a_sample_fmt;
#endif
        
        st->time_base = {1, setting.a_sample_rate};

        AVCodecContext *codec_ctx = ffmpeg_get_codec_context(st);
        codec_ctx->bit_rate = setting.a_bit_rate;
        ffmpeg_set_codec_channels(codec_ctx, setting.a_nb_channel);
        codec_ctx->sample_rate = setting.a_sample_rate;
        codec_ctx->sample_fmt = setting.a_sample_fmt;
        codec_ctx->time_base = {1, setting.a_sample_rate};
        ffmpeg_set_codec_channel_layout(codec_ctx, ffmpeg_get_default_channel_layout(setting.a_nb_channel));

        // 检查是否为RTMP推流（FLV格式）
        bool isRtmpStream = (_fmt_ctx->oformat && strcmp(_fmt_ctx->oformat->name, "flv") == 0);
        
        if (_fmt_ctx->oformat->flags & AVFMT_GLOBALHEADER) {
            if (isRtmpStream) {
                // RTMP推流不使用GLOBAL_HEADER，保持ADTS格式
                al_debug("RTMP stream detected, not setting GLOBAL_HEADER for AAC");
            } else {
                // 其他格式使用GLOBAL_HEADER
                codec_ctx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
                
                codec_ctx->extradata_size
                    = _a_stream->a_enc->get_extradata_size();
                codec_ctx->extradata = (uint8_t *) av_memdup(_a_stream->a_enc->get_extradata(),
                                                             _a_stream->a_enc->get_extradata_size());
                al_debug("Non-RTMP stream, setting GLOBAL_HEADER for AAC");
            }
        }

        // 始终为音频流设置extradata（AudioSpecificConfig），RTSP/SDP/MP4/FLV等容器需要该信息
        if (_a_stream->a_enc->get_extradata_size() > 0) {
            uint8_t *asc = (uint8_t *)av_memdup(_a_stream->a_enc->get_extradata(),
                                                _a_stream->a_enc->get_extradata_size());
            ffmpeg_set_stream_extradata(st, asc, _a_stream->a_enc->get_extradata_size());
            al_debug("Set AAC extradata on stream: size=%d", _a_stream->a_enc->get_extradata_size());
        } else {
            al_warn("AAC extradata size is 0; some outputs (e.g., RTSP/MP4) may fail in avformat_write_header");
        }

        _a_stream->st = st;

        _a_stream->setting = setting;
        // 统一关闭 AAC 比特流过滤器：编码器输出原始 AAC（无 ADTS），由容器负责正确封装
        _a_stream->filter = nullptr;
        al_debug("AAC bitstream filter disabled; container will handle framing based on extradata");

        } while (0);

        return error;
    }

    int muxer_ffmpeg::open_output(const char *output_file, const MUX_SETTING_T &setting)
    {
        int error = AE_NO;
        int ret = 0;

        do {
            if (!(_fmt->flags & AVFMT_NOFILE)) {
                ret = avio_open(&_fmt_ctx->pb, output_file, AVIO_FLAG_WRITE);
                if (ret < 0) {
                error = AE_FFMPEG_OPEN_IO_FAILED;
                break;
            }
        }

        AVDictionary *opt = NULL;
        
        // 检查是否为RTSP推流，添加特定参数
        std::string url_str(output_file);
        if (url_str.find("rtsp://") == 0) {
            // RTSP推流超低延迟参数设置
            av_dict_set(&opt, "fflags", "+genpts+flush_packets+nobuffer", 0); // 生成PTS、立即刷新包、无缓冲
            av_dict_set(&opt, "max_delay", "0", 0);         // 最小延迟
            av_dict_set(&opt, "tune", "zerolatency", 0);    // 零延迟调优
            av_dict_set(&opt, "buffer_size", "512000", 0);  // 减小缓冲区到512KB
            av_dict_set(&opt, "max_interleave_delta", "0", 0); // 最小交错延迟
            // 与播放器保持一致，优先使用TCP传输，提升稳定性并与解码端设置对齐
            av_dict_set(&opt, "rtsp_transport", "tcp", 0);
            al_debug("RTSP output detected, setting ultra-low-latency parameters");
        } else if (url_str.find("rtmp://") == 0) {
            // RTMP推流超低延迟参数设置
            av_dict_set(&opt, "muxdelay", "0", 0);
            av_dict_set(&opt, "fflags", "+genpts+flush_packets+nobuffer", 0);
            av_dict_set(&opt, "max_delay", "0", 0);
            av_dict_set(&opt, "buffer_size", "512000", 0);  // 减小缓冲区到512KB
            av_dict_set(&opt, "max_interleave_delta", "0", 0);
            al_debug("RTMP output detected, setting ultra-low-latency parameters");
        } else {
            // 非RTSP推流的原有参数
            av_dict_set_int(&opt, "video_track_timescale", _v_stream->setting.v_frame_rate, 0);
        }

        ret = avformat_write_header(_fmt_ctx, &opt);

        av_dict_free(&opt);

        if (ret < 0) {
            al_debug("avformat_write_header failed with ret=%d, error=%s", ret, av_err2str(ret));
            error = AE_FFMPEG_WRITE_HEADER_FAILED;
            break;
        }
    } while (0);

    return error;
}

void muxer_ffmpeg::cleanup_video()
{
    if (!_v_stream)
        return;

    if (_v_stream->v_enc)
        delete _v_stream->v_enc;

    if (_v_stream->v_sws)
        delete _v_stream->v_sws;

    delete _v_stream;

    _v_stream = nullptr;
}

void muxer_ffmpeg::cleanup_audio()
{
    if (!_a_stream)
        return;

    if (_a_stream->a_enc)
        delete _a_stream->a_enc;

    if (_a_stream->a_filter_amix)
        delete _a_stream->a_filter_amix;
        
    // 释放bitstream过滤器
    if (_a_stream->filter) {
        ffmpeg_bitstream_filter_close(_a_stream->filter);
        _a_stream->filter = nullptr;
    }

    if (_a_stream->a_nb) {
        for (int i = 0; i < _a_stream->a_nb; i++) {
            if (_a_stream->a_filter_aresample && _a_stream->a_filter_aresample[i])
                delete _a_stream->a_filter_aresample[i];

            if (_a_stream->a_samples && _a_stream->a_samples[i]) {
                delete[] _a_stream->a_samples[i]->buff;
                delete _a_stream->a_samples[i];
            }

            if (_a_stream->a_resamples && _a_stream->a_resamples[i]) {
                delete[] _a_stream->a_resamples[i]->buff;
                delete _a_stream->a_resamples[i];
            }
        }

        if (_a_stream->a_filter_aresample)
            delete[] _a_stream->a_filter_aresample;

        if (_a_stream->a_samples)
            delete[] _a_stream->a_samples;

        if (_a_stream->a_resamples)
            delete[] _a_stream->a_resamples;
    }

    delete _a_stream;

    _a_stream = nullptr;
}

void muxer_ffmpeg::cleanup()
{
    cleanup_video();
    cleanup_audio();

    if (_fmt && !(_fmt->flags & AVFMT_NOFILE))
        avio_closep(&_fmt_ctx->pb);

    if (_fmt_ctx) {
        avformat_free_context(_fmt_ctx);
    }

    _fmt_ctx = NULL;
    _fmt = NULL;

    _inited = false;
}

uint64_t muxer_ffmpeg::get_current_time()
{
    std::lock_guard<std::mutex> lock(_time_mutex);

    return av_gettime_relative();
}

int muxer_ffmpeg::write_video(AVPacket *packet)
{
    //must lock here,coz av_interleaved_write_frame will push packet into a queue,and is not thread safe
    std::lock_guard<std::mutex> lock(_mutex);

    packet->stream_index = _v_stream->st->index;

    // scale ts with timebase of base_time
    av_packet_rescale_ts(packet, _v_stream->v_enc->get_time_base(), {1, AV_TIME_BASE});

    // make audio and video use one clock
    packet->pts = packet->pts - _base_time;
    packet->dts = packet->pts; //make sure that dts is equal to pts

    av_packet_rescale_ts(packet, {1, AV_TIME_BASE}, _v_stream->st->time_base);

    al_debug("V:%lld", packet->pts);

    av_assert0(packet->data != NULL);

    int ret = av_interleaved_write_frame(_fmt_ctx,
                                         packet); //no need to unref packet,this will be auto unref

    if (ret != 0) {
        al_fatal("write video frame error:%d", ret);
    }

    return ret;
}

int muxer_ffmpeg::write_audio(AVPacket *packet)
{
    //must lock here,coz av_interleaved_write_frame will push packet into a queue,and is not thread safe
    std::lock_guard<std::mutex> lock(_mutex);

    packet->stream_index = _a_stream->st->index;

    if (packet->pts == AV_NOPTS_VALUE) {
        packet->pts = av_gettime_relative();
    }

    // scale ts with timebase of base_time
    av_packet_rescale_ts(packet, _a_stream->a_enc->get_time_base(), {1, AV_TIME_BASE});

    // make audio and video use one clock
    packet->pts = packet->pts - _base_time;
    packet->dts = packet->pts; //make sure that dts is equal to pts

    av_packet_rescale_ts(packet, {1, AV_TIME_BASE}, _a_stream->st->time_base);

    al_debug("A:%lld %lld", packet->pts, packet->dts);

    av_assert0(packet->data != NULL);

    int ret = av_interleaved_write_frame(_fmt_ctx,
                                         packet); //no need to unref packet,this will be auto unref

    if (ret != 0) {
        al_fatal("write audio frame error:%d", ret);
    }

    return ret;
}
} // namespace am