ffmpeg取rtsp流音频数据保存声音为wav文件

本来不是什么难搞的问题，代码写完了，音频流信息中的详细信息，具体代码表现为

format_ctx->streams[audio_stream_index]->codecpar是空指针。

这个查了一圈也没人给出正确答案，实际上是由于我自己编译的ffmpeg时候，开启的选项的导致的。把音频解码器相关的给禁掉了。重新开启相关编译选项，编译ffmpeg后，一切正常。

具体的选项为：

ffmpeg 交叉编译./configure --prefix=../arm-ffmpegbuild \
--enable-shared \
--enable-libmp3lame \--enable-libx264 \--enable-gpl \--disable-asm \--enable-version3 \--enable-libmp3lame \--enable-libx264 \--enable-libvpx \--enable-nonfree \--cross-prefix=aarch64-linux- \--target-os=linux \--extra-cflags="-I /opt/ffmpeg_test_make/lame-3.100/lamebuild/include" \--extra-ldflags="-L /opt/ffmpeg_test_make/lame-3.100/lamebuild/lib" \--enable-cross-compile \--enable-small \--arch=arm64 \--enable-decoder=h264 \--enable-parser=h264 \--enable-demuxer=rtsp \--extra-ldflags="-L ../x264build/lib" \--extra-cflags="-I ../x264build/include"lame交叉编译./configure \--host=aarch64-linux \--prefix=/opt/ffmpeg_test_make/lame-3.100/lamebuild \cc=aarch64-linux-gcc 

话不多说上代码：

bool FfpDecoderWav::dump_wav(std::string rtsp_url, std::string file_path) {AVDictionary *format_options = NULL;av_dict_set(&format_options, "rtsp_transport", "tcp", 0); // 以tcp的方式打开,av_register_all();avformat_network_init();// 打开 RTSP 流int reconnect_times = 3;AVFormatContext *format_ctx = NULL;bool online = false;while (reconnect_times-- > 0) {if (format_ctx != NULL) {avformat_close_input(&format_ctx);format_ctx = NULL;}format_ctx = avformat_alloc_context();if (avformat_open_input(&format_ctx, rtsp_url.c_str(), NULL, &format_options) != 0) {Logger::error("open rtsp url:{} faile", rtsp_url);// std::this_thread::sleep_for(std::chrono::milliseconds(500));usleep(100000);} else {online = true;break;}}av_dict_free(&format_options); // 释放 format_optionsif (!online) {return false;}Logger::info("open rtsp url:{} for wav success", rtsp_url);// 查找音频流int audio_stream_index = -1;if (avformat_find_stream_info(format_ctx, NULL) < 0) {Logger::info("can not avformat_find_stream_info url:{}", rtsp_url);return false;}AVCodec *codec = NULL;audio_stream_index = av_find_best_stream(format_ctx, AVMEDIA_TYPE_AUDIO, -1, -1, &codec, 0);std::cout << "codec name :" << codec->name << std::endl;std::cout << "codec long_name :" << codec->long_name << std::endl;std::cout << "codec AVMediaType :" << (int)codec->type << std::endl;std::cout << "codec AVCodecID :" << (int)codec->id << std::endl;if (audio_stream_index < 0 || codec == NULL) {Logger::info("can not find sound stream rtsp url:{}", rtsp_url);return false;}Logger::info("find sound stream success index:{}", audio_stream_index);av_dump_format(format_ctx, 0, rtsp_url.c_str(), 0);bool had_audio_code = true;SwrContext *swr_ctx = NULL;AVCodecContext *codec_ctx = avcodec_alloc_context3(codec);if (format_ctx->streams[audio_stream_index]->codecpar) {Logger::info("avcodec_alloc_context3 success channels={}", codec_ctx->channels);Logger::info("avcodec_alloc_context3 success sample_rate={}", codec_ctx->sample_rate);// std::cout << "had codecpar inf" << std::endl;// printf("had codecpar inf\n");avcodec_parameters_to_context(codec_ctx, format_ctx->streams[audio_stream_index]->codecpar);if (avcodec_open2(codec_ctx, codec, NULL) < 0) {Logger::info("avcodec_open2 error rtsp url:{}", rtsp_url);return false;}Logger::info("avcodec_open2 success channels={}", codec_ctx->channels);Logger::info("avcodec_open2 success sample_rate={}", codec_ctx->sample_rate);// 创建重采样上下文swr_ctx = swr_alloc_set_opts(NULL, NUM_CHANNELS, AV_SAMPLE_FMT_S16, SAMPLE_RATE, codec_ctx->channels,codec_ctx->sample_fmt, codec_ctx->sample_rate, 0, NULL);Logger::info("swr_alloc_set_opts success");if (!swr_ctx || swr_init(swr_ctx) < 0) {// Logger::info("swr_init error rtsp url:{}", rtsp_url);return false;}} else {printf("cdecpar is nullodecpar is nullodecpar is nullodecpar is null\n");std::cout << "codecpar is null" << std::endl;had_audio_code = false;}// 创建输出 WAV 文件std::ofstream wav_file(file_path.c_str(), std::ios::binary);if (!wav_file) {// Logger::info("fopen local_path save wav failed path:{}", file_path);return false;}// Logger::info("open wav_file success");//  写入 WAV 文件头WAVHeader wav_header;unsigned int file_size = sizeof(wav_header);// Logger::info("wav_header size:{}", file_size);wav_file.write((const char *)&wav_header, file_size);time_t start_time = time(NULL);AVPacket packet;int ret = 0;int count = 1000;while (true) {if (ret = av_read_frame(format_ctx, &packet) < 0) {// Logger::info("av_read_frame failed: {}", ret);break;}time_t current_time = time(NULL);time_t duration = current_time - start_time;if (duration > 60) {// Logger::info("save sound end by 20 s time");break;}if (packet.stream_index == audio_stream_index) {if (!had_audio_code) {wav_file.write((char *)packet.data, packet.size);std::cout << "write sws data codecpar inf insfsjfjaslkjfas" << std::endl;printf(" wav_file.write((char *)packet.data, packet.size);\n");continue;}AVFrame *frame = av_frame_alloc();if (avcodec_send_packet(codec_ctx, &packet) >= 0 && avcodec_receive_frame(codec_ctx, frame) >= 0) {uint8_t *out_buffer[NUM_CHANNELS];int out_samples = 0;int out_size = 0;for (int i = 0; i < NUM_CHANNELS; i++) {out_buffer[i] = (uint8_t *)malloc(frame->nb_samples * 2 * sizeof(uint8_t));}out_samples = swr_convert(swr_ctx, out_buffer, frame->nb_samples, (const uint8_t **)frame->data,frame->nb_samples);out_size = out_samples * NUM_CHANNELS * 2;wav_file.write(reinterpret_cast<char *>(out_buffer[0]), out_size);// std::cout << "write sws data codecpar inf" << std::endl;// printf(" wav_file.write(reinterpret_cast<char *>(out_buffer[0]), out_siz22;\n");for (int i = 0; i < NUM_CHANNELS; i++) {free(out_buffer[i]);}}av_frame_free(&frame);}av_packet_unref(&packet);}// 更新 WAV 文件头中的数据大小uint32_t subchunk2Size = static_cast<unsigned int>(wav_file.tellp()) - 44;uint32_t chunkSize = subchunk2Size + 36;wav_file.seekp(4, std::ios::beg);wav_file.write(reinterpret_cast<char *>(&chunkSize), 4);wav_file.seekp(40, std::ios::beg);wav_file.write(reinterpret_cast<char *>(&subchunk2Size), 4);// 关闭文件wav_file.close();// 释放资源avcodec_close(codec_ctx);avcodec_free_context(&codec_ctx);avformat_close_input(&format_ctx);swr_free(&swr_ctx);// Logger::info("save local_path  wav success path:{}", file_path);return true;
}

wav格式的数据头文件：

struct WAVHeader {char chunkID[4] = {'R', 'I', 'F', 'F'};uint32_t chunkSize = 0;char format[4] = {'W', 'A', 'V', 'E'};char subchunk1ID[4] = {'f', 'm', 't', ' '};uint32_t subchunk1Size = 16;uint16_t audioFormat = 1;uint16_t numChannels = NUM_CHANNELS;uint32_t sampleRate = SAMPLE_RATE;uint32_t byteRate = SAMPLE_RATE * NUM_CHANNELS * 16 / 8;uint16_t blockAlign = 4;uint16_t bitsPerSample = 16;char subchunk2ID[4] = {'d', 'a', 't', 'a'};uint32_t subchunk2Size = 4;
};

最后，就是wav注意的地方，一共是两个值：

chunkSize 和subchunk2Size

// 更新 WAV 文件头中的数据大小

也就是说：subchunk2Size是出去wav文件头部数据意外的数据长度。

即文件总长度减去头部长度44个字节。

chunkSize=subchunk2Size+36

具体为什么，可以查看wav格式的说明。