Make it possible for file and HTTP streams to use different audio codecs.

diff --git a/flags.cpp b/flags.cpp
index 0c01e23799812a3a4f0060ac33d5bb98cbf12b55..398d49d5bffd65465e2227e13c1df5f2eba5d242 100644 (file)
--- a/flags.cpp
+++ b/flags.cpp
@@ -17,6 +17,11 @@ void usage()
        fprintf(stderr, "                                    ($DISPLAY spec or /dev/dri/render* path)\n");
        fprintf(stderr, "      --http-uncompressed-video   send uncompressed NV12 video to HTTP clients\n");
        fprintf(stderr, "      --http-mux=NAME             mux to use for HTTP streams (default " DEFAULT_STREAM_MUX_NAME ")\n");
+       fprintf(stderr, "      --http-audio-codec=NAME     audio codec to use for HTTP streams\n");
+       fprintf(stderr, "                                  (default is to use the same as for the recording)\n");
+       fprintf(stderr, "      --http-audio-bitrate=KBITS  audio codec bit rate to use for HTTP streams\n");
+       fprintf(stderr, "                                  (default is %d, ignored unless --http-audio-codec is set)\n",
+               DEFAULT_AUDIO_OUTPUT_BIT_RATE / 1000);
        fprintf(stderr, "      --http-coarse-timebase      use less timebase for HTTP (recommended for muxers\n");
        fprintf(stderr, "                                  that handle large pts poorly, like e.g. MP4)\n");
        fprintf(stderr, "      --flat-audio                start with most audio processing turned off\n");
@@ -35,6 +40,8 @@ void parse_flags(int argc, char * const argv[])
                { "http-uncompressed-video", no_argument, 0, 1001 },
                { "http-mux", required_argument, 0, 1004 },
                { "http-coarse-timebase", no_argument, 0, 1005 },
+               { "http-audio-codec", required_argument, 0, 1006 },
+               { "http-audio-bitrate", required_argument, 0, 1007 },
                { "flat-audio", no_argument, 0, 1002 },
                { "no-flush-pbos", no_argument, 0, 1003 },
                { 0, 0, 0, 0 }
@@ -65,6 +72,12 @@ void parse_flags(int argc, char * const argv[])
                case 1005:
                        global_flags.stream_coarse_timebase = true;
                        break;
+               case 1006:
+                       global_flags.stream_audio_codec_name = optarg;
+                       break;
+               case 1007:
+                       global_flags.stream_audio_codec_bitrate = atoi(optarg) * 1000;
+                       break;
                case 1002:
                        global_flags.flat_audio = true;
                        break;

diff --git a/flags.h b/flags.h
index b2dedb6c05a1ca689b1d5b588b994142da552f20..7dc03c546fdce57079aa3d2203175bbdbd70edc9 100644 (file)
--- a/flags.h
+++ b/flags.h
@@ -14,6 +14,8 @@ struct Flags {
        bool flush_pbos = true;
        std::string stream_mux_name = DEFAULT_STREAM_MUX_NAME;
        bool stream_coarse_timebase = false;
+       std::string stream_audio_codec_name;  // Blank = use the same as for the recording.
+       int stream_audio_codec_bitrate = DEFAULT_AUDIO_OUTPUT_BIT_RATE;  // Ignored if stream_audio_codec_name is blank.
 };
 extern Flags global_flags;
 

diff --git a/h264encode.cpp b/h264encode.cpp
index 3389a1ebecd35d5b40a812d1241350ebbffa715d..8f4508835c0a2f3d57a02c5e5c52f258f73344c2 100644 (file)
--- a/h264encode.cpp
+++ b/h264encode.cpp
@@ -228,11 +228,13 @@ private:
        void encode_audio(const vector<float> &audio,
                          vector<float> *audio_queue,
                          int64_t audio_pts,
-                         AVCodecContext *ctx);
+                         AVCodecContext *ctx,
+                         const vector<PacketDestination *> &destinations);
        void encode_audio_one_frame(const float *audio,
                                    size_t num_samples,  // In each channel.
                                    int64_t audio_pts,
-                                   AVCodecContext *ctx);
+                                   AVCodecContext *ctx,
+                                   const vector<PacketDestination *> &destinations);
        void storage_task_enqueue(storage_task task);
        void save_codeddata(storage_task task);
        int render_packedsequence();
@@ -279,8 +281,11 @@ private:
        map<int64_t, vector<float>> pending_audio_frames;  // under frame_queue_mutex
        QSurface *surface;
 
-       AVCodecContext *context_audio;
-       vector<float> audio_queue;
+       AVCodecContext *context_audio_file;
+       AVCodecContext *context_audio_stream = nullptr;  // nullptr = don't code separate audio for stream.
+
+       vector<float> audio_queue_file;
+       vector<float> audio_queue_stream;
 
        AVFrame *audio_frame = nullptr;
        HTTPD *httpd;
@@ -1653,7 +1658,12 @@ void H264EncoderImpl::save_codeddata(storage_task task)
                        pending_audio_frames.erase(it); 
                }
 
-               encode_audio(audio, &audio_queue, audio_pts, context_audio);
+               if (context_audio_stream) {
+                       encode_audio(audio, &audio_queue_file, audio_pts, context_audio_file, { file_mux.get() });
+                       encode_audio(audio, &audio_queue_stream, audio_pts, context_audio_stream, { httpd });
+               } else {
+                       encode_audio(audio, &audio_queue_file, audio_pts, context_audio_file, { httpd, file_mux.get() });
+               }
 
                if (audio_pts == task.pts) break;
        }
@@ -1663,13 +1673,14 @@ void H264EncoderImpl::encode_audio(
        const vector<float> &audio,
        vector<float> *audio_queue,
        int64_t audio_pts,
-       AVCodecContext *ctx)
+       AVCodecContext *ctx,
+       const vector<PacketDestination *> &destinations)
 {
        if (ctx->frame_size == 0) {
                // No queueing needed.
                assert(audio_queue->empty());
                assert(audio.size() % 2 == 0);
-               encode_audio_one_frame(&audio[0], audio.size() / 2, audio_pts, ctx);
+               encode_audio_one_frame(&audio[0], audio.size() / 2, audio_pts, ctx, destinations);
                return;
        }
 
@@ -1681,7 +1692,8 @@ void H264EncoderImpl::encode_audio(
                encode_audio_one_frame(&(*audio_queue)[sample_num],
                                       ctx->frame_size,
                                       audio_pts,
-                                      ctx);
+                                      ctx,
+                                      destinations);
        }
        audio_queue->erase(audio_queue->begin(), audio_queue->begin() + sample_num);
 }
@@ -1690,7 +1702,8 @@ void H264EncoderImpl::encode_audio_one_frame(
        const float *audio,
        size_t num_samples,
        int64_t audio_pts,
-       AVCodecContext *ctx)
+       AVCodecContext *ctx,
+       const vector<PacketDestination *> &destinations)
 {
        audio_frame->nb_samples = num_samples;
        audio_frame->channel_layout = AV_CH_LAYOUT_STEREO;
@@ -1734,10 +1747,9 @@ void H264EncoderImpl::encode_audio_one_frame(
        if (got_output) {
                pkt.stream_index = 1;
                pkt.flags = AV_PKT_FLAG_KEY;
-               if (file_mux) {
-                       file_mux->add_packet(pkt, audio_pts + global_delay(), audio_pts + global_delay());
+               for (PacketDestination *dest : destinations) {
+                       dest->add_packet(pkt, audio_pts + global_delay(), audio_pts + global_delay());
                }
-               httpd->add_packet(pkt, audio_pts + global_delay(), audio_pts + global_delay());
        }
        // TODO: Delayed frames.
        av_frame_unref(audio_frame);
@@ -1856,7 +1868,12 @@ void init_audio_encoder(const string &codec_name, int bit_rate, AVCodecContext *
 H264EncoderImpl::H264EncoderImpl(QSurface *surface, const string &va_display, int width, int height, HTTPD *httpd)
        : current_storage_frame(0), surface(surface), httpd(httpd)
 {
-       init_audio_encoder(AUDIO_OUTPUT_CODEC_NAME, AUDIO_OUTPUT_BIT_RATE, &context_audio);
+       init_audio_encoder(AUDIO_OUTPUT_CODEC_NAME, DEFAULT_AUDIO_OUTPUT_BIT_RATE, &context_audio_file);
+
+       if (!global_flags.stream_audio_codec_name.empty()) {
+               init_audio_encoder(global_flags.stream_audio_codec_name,
+                       global_flags.stream_audio_codec_bitrate, &context_audio_stream);
+       }
 
        audio_frame = av_frame_alloc();
 
@@ -2071,7 +2088,7 @@ void H264EncoderImpl::open_output_file(const std::string &filename)
                exit(1);
        }
 
-       file_mux.reset(new Mux(avctx, frame_width, frame_height, Mux::CODEC_H264, TIMEBASE));
+       file_mux.reset(new Mux(avctx, frame_width, frame_height, Mux::CODEC_H264, TIMEBASE, DEFAULT_AUDIO_OUTPUT_BIT_RATE));
 }
 
 void H264EncoderImpl::close_output_file()

diff --git a/httpd.cpp b/httpd.cpp
index 6bf436478b8df8cd9afb83301b0ec6a330f808c7..487d429c5d5b1e49df0a4e019f954823ffb2ceee 100644 (file)
--- a/httpd.cpp
+++ b/httpd.cpp
@@ -67,8 +67,13 @@ int HTTPD::answer_to_connection(MHD_Connection *connection,
        AVOutputFormat *oformat = av_guess_format(global_flags.stream_mux_name.c_str(), nullptr, nullptr);
        assert(oformat != nullptr);
 
+       // TODO: This is an ugly place to have this logic.
+       const int bit_rate = global_flags.stream_audio_codec_name.empty() ?
+               DEFAULT_AUDIO_OUTPUT_BIT_RATE :
+               global_flags.stream_audio_codec_bitrate;
+
        int time_base = global_flags.stream_coarse_timebase ? COARSE_TIMEBASE : TIMEBASE;
-       HTTPD::Stream *stream = new HTTPD::Stream(oformat, width, height, time_base);
+       HTTPD::Stream *stream = new HTTPD::Stream(oformat, width, height, time_base, bit_rate);
        {
                unique_lock<mutex> lock(streams_mutex);
                streams.insert(stream);
@@ -112,7 +117,7 @@ void HTTPD::request_completed(struct MHD_Connection *connection, void **con_cls,
        }
 }
 
-HTTPD::Stream::Stream(AVOutputFormat *oformat, int width, int height, int time_base)
+HTTPD::Stream::Stream(AVOutputFormat *oformat, int width, int height, int time_base, int bit_rate)
 {
        AVFormatContext *avctx = avformat_alloc_context();
        avctx->oformat = oformat;
@@ -128,7 +133,7 @@ HTTPD::Stream::Stream(AVOutputFormat *oformat, int width, int height, int time_b
 
        avctx->flags = AVFMT_FLAG_CUSTOM_IO;
 
-       mux.reset(new Mux(avctx, width, height, video_codec, time_base));
+       mux.reset(new Mux(avctx, width, height, video_codec, time_base, bit_rate));
 }
 
 ssize_t HTTPD::Stream::reader_callback_thunk(void *cls, uint64_t pos, char *buf, size_t max)

diff --git a/httpd.h b/httpd.h
index 55a81c20e59ffa125ef59a28320e98aa0f34c494..11f15bd81fd9314433edf72e4da19d6e6e2bcb39 100644 (file)
--- a/httpd.h
+++ b/httpd.h
@@ -28,11 +28,11 @@ extern "C" {
 
 #include "mux.h"
 
-class HTTPD {
+class HTTPD : public PacketDestination {
 public:
        HTTPD(int width, int height);
        void start(int port);
-       void add_packet(const AVPacket &pkt, int64_t pts, int64_t dts);
+       void add_packet(const AVPacket &pkt, int64_t pts, int64_t dts) override;
 
 private:
        static int answer_to_connection_thunk(void *cls, MHD_Connection *connection,
@@ -54,7 +54,7 @@ private:
 
        class Stream {
        public:
-               Stream(AVOutputFormat *oformat, int width, int height, int time_base);
+               Stream(AVOutputFormat *oformat, int width, int height, int time_base, int bit_rate);
 
                static ssize_t reader_callback_thunk(void *cls, uint64_t pos, char *buf, size_t max);
                ssize_t reader_callback(uint64_t pos, char *buf, size_t max);

diff --git a/mux.cpp b/mux.cpp
index 82d42abe705c75e2af79259252a68e5494e6763f..3ff787f65a96dfe87e3b4230ce5fcd9dd678b635 100644 (file)
--- a/mux.cpp
+++ b/mux.cpp
@@ -9,7 +9,7 @@
 
 using namespace std;
 
-Mux::Mux(AVFormatContext *avctx, int width, int height, Codec video_codec, int time_base)
+Mux::Mux(AVFormatContext *avctx, int width, int height, Codec video_codec, int time_base, int bit_rate)
        : avctx(avctx)
 {
        AVCodec *codec_video = avcodec_find_encoder((video_codec == CODEC_H264) ? AV_CODEC_ID_H264 : AV_CODEC_ID_RAWVIDEO);
@@ -57,7 +57,7 @@ Mux::Mux(AVFormatContext *avctx, int width, int height, Codec video_codec, int t
                exit(1);
        }
        avstream_audio->time_base = AVRational{1, time_base};
-       avstream_audio->codec->bit_rate = AUDIO_OUTPUT_BIT_RATE;
+       avstream_audio->codec->bit_rate = bit_rate;
        avstream_audio->codec->sample_rate = OUTPUT_FREQUENCY;
        avstream_audio->codec->channels = 2;
        avstream_audio->codec->channel_layout = AV_CH_LAYOUT_STEREO;

diff --git a/mux.h b/mux.h
index 2aefffcfeae704736a2ab1b8b3fd25593d1b1756..d818bdab8b8b92e24b481ea88cfcdba0d53f4cbc 100644 (file)
--- a/mux.h
+++ b/mux.h
@@ -9,16 +9,23 @@ extern "C" {
 #include <libavformat/avio.h>
 }
 
-class Mux {
+class PacketDestination {
+public:
+       virtual ~PacketDestination() {}
+       virtual void add_packet(const AVPacket &pkt, int64_t pts, int64_t dts) = 0;
+};
+
+class Mux : public PacketDestination {
 public:
        enum Codec {
                CODEC_H264,
                CODEC_NV12,  // Uncompressed 4:2:0.
        };
 
-       Mux(AVFormatContext *avctx, int width, int height, Codec video_codec, int time_base);  // Takes ownership of avctx.
+       // Takes ownership of avctx.
+       Mux(AVFormatContext *avctx, int width, int height, Codec video_codec, int time_base, int bit_rate);
        ~Mux();
-       void add_packet(const AVPacket &pkt, int64_t pts, int64_t dts);
+       void add_packet(const AVPacket &pkt, int64_t pts, int64_t dts) override;
 
 private:
        bool seen_keyframe = false;