Add support for AAC audio codec

Add option --audio-codec=aac.

app/meson.build

@@ -21,6 +21,7 @@ src = [
     'src/mouse_inject.c',
     'src/opengl.c',
     'src/options.c',
+    'src/packet_merger.c',
     'src/receiver.c',
     'src/recorder.c',
     'src/scrcpy.c',
@@ -200,10 +201,6 @@ conf.set('PORTABLE', get_option('portable'))
 conf.set('DEFAULT_LOCAL_PORT_RANGE_FIRST', '27183')
 conf.set('DEFAULT_LOCAL_PORT_RANGE_LAST', '27199')
 
-# the default video bitrate, in bits/second
-# overridden by option --bit-rate
-conf.set('DEFAULT_BIT_RATE', '8000000')  # 8Mbps
-
 # run a server debugger and wait for a client to be attached
 conf.set('SERVER_DEBUGGER', get_option('server_debugger'))
 

app/scrcpy.1

@@ -19,16 +19,30 @@ provides display and control of Android devices connected on USB (or over TCP/IP
 .B \-\-always\-on\-top
 Make scrcpy window always on top (above other windows).
 
+.TP
+.BI "\-\-audio\-bit\-rate " value
+Encode the audio at the given bit\-rate, expressed in bits/s. Unit suffixes are supported: '\fBK\fR' (x1000) and '\fBM\fR' (x1000000).
+
+Default is 196K (196000).
+
+.TP
+.BI "\-\-audio\-codec " name
+Select an audio codec (opus or aac).
+
+Default is opus.
+
 .TP
 .BI "\-b, \-\-bit\-rate " value
 Encode the video at the given bit\-rate, expressed in bits/s. Unit suffixes are supported: '\fBK\fR' (x1000) and '\fBM\fR' (x1000000).
 
-Default is 8000000.
+Default is 8M (8000000).
 
 .TP
 .BI "\-\-codec " name
 Select a video codec (h264, h265 or av1).
 
+Default is h264.
+
 .TP
 .BI "\-\-codec\-options " key\fR[:\fItype\fR]=\fIvalue\fR[,...]
 Set a list of comma-separated key:type=value options for the device encoder.

app/src/cli.c

@@ -59,6 +59,8 @@
 #define OPT_NO_POWER_ON            1039
 #define OPT_CODEC                  1040
 #define OPT_NO_AUDIO               1041
+#define OPT_AUDIO_BIT_RATE         1042
+#define OPT_AUDIO_CODEC            1043
 
 struct sc_option {
     char shortopt;
@@ -99,19 +101,35 @@ static const struct sc_option options[] = {
         .longopt = "always-on-top",
         .text = "Make scrcpy window always on top (above other windows).",
     },
+    {
+        .longopt_id = OPT_AUDIO_BIT_RATE,
+        .longopt = "audio-bit-rate",
+        .argdesc = "value",
+        .text = "Encode the audio at the given bit-rate, expressed in bits/s. "
+                "Unit suffixes are supported: 'K' (x1000) and 'M' (x1000000).\n"
+                "Default is 196K (196000).",
+    },
+    {
+        .longopt_id = OPT_AUDIO_CODEC,
+        .longopt = "audio-codec",
+        .argdesc = "name",
+        .text = "Select an audio codec (opus or aac).\n"
+                "Default is opus.",
+    },
     {
         .shortopt = 'b',
         .longopt = "bit-rate",
         .argdesc = "value",
-        .text = "Encode the video at the gitven bit-rate, expressed in bits/s. "
+        .text = "Encode the video at the given bit-rate, expressed in bits/s. "
                 "Unit suffixes are supported: 'K' (x1000) and 'M' (x1000000).\n"
-                "Default is " STR(DEFAULT_BIT_RATE) ".",
+                "Default is 8M (8000000).",
     },
     {
         .longopt_id = OPT_CODEC,
         .longopt = "codec",
         .argdesc = "name",
-        .text = "Select a video codec (h264, h265 or av1).",
+        .text = "Select a video codec (h264, h265 or av1).\n"
+                "Default is h264.",
     },
     {
         .longopt_id = OPT_CODEC_OPTIONS,
@@ -1408,6 +1426,20 @@ parse_codec(const char *optarg, enum sc_codec *codec) {
     return false;
 }
 
+static bool
+parse_audio_codec(const char *optarg, enum sc_codec *codec) {
+    if (!strcmp(optarg, "opus")) {
+        *codec = SC_CODEC_OPUS;
+        return true;
+    }
+    if (!strcmp(optarg, "aac")) {
+        *codec = SC_CODEC_AAC;
+        return true;
+    }
+    LOGE("Unsupported audio codec: %s (expected opus)", optarg);
+    return false;
+}
+
 static bool
 parse_args_with_getopt(struct scrcpy_cli_args *args, int argc, char *argv[],
                        const char *optstring, const struct option *longopts) {
@@ -1423,6 +1455,11 @@ parse_args_with_getopt(struct scrcpy_cli_args *args, int argc, char *argv[],
                     return false;
                 }
                 break;
+            case OPT_AUDIO_BIT_RATE:
+                if (!parse_bit_rate(optarg, &opts->audio_bit_rate)) {
+                    return false;
+                }
+                break;
             case OPT_CROP:
                 opts->crop = optarg;
                 break;
@@ -1649,6 +1686,11 @@ parse_args_with_getopt(struct scrcpy_cli_args *args, int argc, char *argv[],
                     return false;
                 }
                 break;
+            case OPT_AUDIO_CODEC:
+                if (!parse_audio_codec(optarg, &opts->audio_codec)) {
+                    return false;
+                }
+                break;
             case OPT_OTG:
 #ifdef HAVE_USB
                 opts->otg = true;

app/src/controller.c

@@ -9,20 +9,20 @@ sc_controller_init(struct sc_controller *controller, sc_socket control_socket,
                    struct sc_acksync *acksync) {
     cbuf_init(&controller->queue);
 
-    bool ok = receiver_init(&controller->receiver, control_socket, acksync);
+    bool ok = sc_receiver_init(&controller->receiver, control_socket, acksync);
     if (!ok) {
         return false;
     }
 
     ok = sc_mutex_init(&controller->mutex);
     if (!ok) {
-        receiver_destroy(&controller->receiver);
+        sc_receiver_destroy(&controller->receiver);
         return false;
     }
 
     ok = sc_cond_init(&controller->msg_cond);
     if (!ok) {
-        receiver_destroy(&controller->receiver);
+        sc_receiver_destroy(&controller->receiver);
         sc_mutex_destroy(&controller->mutex);
         return false;
     }
@@ -43,7 +43,7 @@ sc_controller_destroy(struct sc_controller *controller) {
         sc_control_msg_destroy(&msg);
     }
 
-    receiver_destroy(&controller->receiver);
+    sc_receiver_destroy(&controller->receiver);
 }
 
 bool
@@ -117,7 +117,7 @@ sc_controller_start(struct sc_controller *controller) {
         return false;
     }
 
-    if (!receiver_start(&controller->receiver)) {
+    if (!sc_receiver_start(&controller->receiver)) {
         sc_controller_stop(controller);
         sc_thread_join(&controller->thread, NULL);
         return false;
@@ -137,5 +137,5 @@ sc_controller_stop(struct sc_controller *controller) {
 void
 sc_controller_join(struct sc_controller *controller) {
     sc_thread_join(&controller->thread, NULL);
-    receiver_join(&controller->receiver);
+    sc_receiver_join(&controller->receiver);
 }

app/src/controller.h

@@ -21,7 +21,7 @@ struct sc_controller {
     sc_cond msg_cond;
     bool stopped;
     struct sc_control_msg_queue queue;
-    struct receiver receiver;
+    struct sc_receiver receiver;
 };
 
 bool

app/src/decoder.c

@@ -29,7 +29,6 @@ sc_decoder_open_sinks(struct sc_decoder *decoder) {
     for (unsigned i = 0; i < decoder->sink_count; ++i) {
         struct sc_frame_sink *sink = decoder->sinks[i];
         if (!sink->ops->open(sink)) {
-            LOGE("Could not open frame sink %d", i);
             sc_decoder_close_first_sinks(decoder, i);
             return false;
         }
@@ -63,7 +62,6 @@ sc_decoder_open(struct sc_decoder *decoder, const AVCodec *codec) {
     }
 
     if (!sc_decoder_open_sinks(decoder)) {
-        LOGE("Could not open decoder sinks");
         av_frame_free(&decoder->frame);
         avcodec_close(decoder->codec_ctx);
         avcodec_free_context(&decoder->codec_ctx);
@@ -86,7 +84,6 @@ push_frame_to_sinks(struct sc_decoder *decoder, const AVFrame *frame) {
     for (unsigned i = 0; i < decoder->sink_count; ++i) {
         struct sc_frame_sink *sink = decoder->sinks[i];
         if (!sink->ops->push(sink, frame)) {
-            LOGE("Could not send frame to sink %d", i);
             return false;
         }
     }

app/src/demuxer.c

@@ -6,6 +6,7 @@
 
 #include "decoder.h"
 #include "events.h"
+#include "packet_merger.h"
 #include "recorder.h"
 #include "util/binary.h"
 #include "util/log.h"
@@ -18,17 +19,12 @@
 #define SC_PACKET_PTS_MASK (SC_PACKET_FLAG_KEY_FRAME - 1)
 
 static enum AVCodecID
-sc_demuxer_recv_codec_id(struct sc_demuxer *demuxer) {
-    uint8_t data[4];
-    ssize_t r = net_recv_all(demuxer->socket, data, 4);
-    if (r < 4) {
-        return false;
-    }
-
+sc_demuxer_to_avcodec_id(uint32_t codec_id) {
 #define SC_CODEC_ID_H264 UINT32_C(0x68323634) // "h264" in ASCII
 #define SC_CODEC_ID_H265 UINT32_C(0x68323635) // "h265" in ASCII
 #define SC_CODEC_ID_AV1 UINT32_C(0x00617631) // "av1" in ASCII
-    uint32_t codec_id = sc_read32be(data);
+#define SC_CODEC_ID_OPUS UINT32_C(0x6f707573) // "opus" in ASCII
+#define SC_CODEC_ID_AAC UINT32_C(0x00616163) // "aac in ASCII"
     switch (codec_id) {
         case SC_CODEC_ID_H264:
             return AV_CODEC_ID_H264;
@@ -36,12 +32,28 @@ sc_demuxer_recv_codec_id(struct sc_demuxer *demuxer) {
             return AV_CODEC_ID_HEVC;
         case SC_CODEC_ID_AV1:
             return AV_CODEC_ID_AV1;
+        case SC_CODEC_ID_OPUS:
+            return AV_CODEC_ID_OPUS;
+        case SC_CODEC_ID_AAC:
+            return AV_CODEC_ID_AAC;
         default:
             LOGE("Unknown codec id 0x%08" PRIx32, codec_id);
             return AV_CODEC_ID_NONE;
     }
 }
 
+static bool
+sc_demuxer_recv_codec_id(struct sc_demuxer *demuxer, uint32_t *codec_id) {
+    uint8_t data[4];
+    ssize_t r = net_recv_all(demuxer->socket, data, 4);
+    if (r < 4) {
+        return false;
+    }
+
+    *codec_id = sc_read32be(data);
+    return true;
+}
+
 static bool
 sc_demuxer_recv_packet(struct sc_demuxer *demuxer, AVPacket *packet) {
     // The video stream contains raw packets, without time information. When we
@@ -105,7 +117,6 @@ push_packet_to_sinks(struct sc_demuxer *demuxer, const AVPacket *packet) {
     for (unsigned i = 0; i < demuxer->sink_count; ++i) {
         struct sc_packet_sink *sink = demuxer->sinks[i];
         if (!sink->ops->push(sink, packet)) {
-            LOGE("Could not send config packet to sink %d", i);
             return false;
         }
     }
@@ -115,50 +126,9 @@ push_packet_to_sinks(struct sc_demuxer *demuxer, const AVPacket *packet) {
 
 static bool
 sc_demuxer_push_packet(struct sc_demuxer *demuxer, AVPacket *packet) {
-    bool is_config = packet->pts == AV_NOPTS_VALUE;
-
-    // A config packet must not be decoded immediately (it contains no
-    // frame); instead, it must be concatenated with the future data packet.
-    if (demuxer->pending || is_config) {
-        if (demuxer->pending) {
-            size_t offset = demuxer->pending->size;
-            if (av_grow_packet(demuxer->pending, packet->size)) {
-                LOG_OOM();
-                return false;
-            }
-
-            memcpy(demuxer->pending->data + offset, packet->data, packet->size);
-        } else {
-            demuxer->pending = av_packet_alloc();
-            if (!demuxer->pending) {
-                LOG_OOM();
-                return false;
-            }
-            if (av_packet_ref(demuxer->pending, packet)) {
-                LOG_OOM();
-                av_packet_free(&demuxer->pending);
-                return false;
-            }
-        }
-
-        if (!is_config) {
-            // prepare the concat packet to send to the decoder
-            demuxer->pending->pts = packet->pts;
-            demuxer->pending->dts = packet->dts;
-            demuxer->pending->flags = packet->flags;
-            packet = demuxer->pending;
-        }
-    }
-
     bool ok = push_packet_to_sinks(demuxer, packet);
-
-    if (!is_config && demuxer->pending) {
-        // the pending packet must be discarded (consumed or error)
-        av_packet_free(&demuxer->pending);
-    }
-
     if (!ok) {
-        LOGE("Could not process packet");
+        LOGE("Demuxer '%s': could not process packet", demuxer->name);
         return false;
     }
 
@@ -183,7 +153,6 @@ sc_demuxer_open_sinks(struct sc_demuxer *demuxer, const AVCodec *codec) {
     for (unsigned i = 0; i < demuxer->sink_count; ++i) {
         struct sc_packet_sink *sink = demuxer->sinks[i];
         if (!sink->ops->open(sink, codec)) {
-            LOGE("Could not open packet sink %d", i);
             sc_demuxer_close_first_sinks(demuxer, i);
             return false;
         }
@@ -192,56 +161,93 @@ sc_demuxer_open_sinks(struct sc_demuxer *demuxer, const AVCodec *codec) {
     return true;
 }
 
+static void
+sc_demuxer_disable_sinks(struct sc_demuxer *demuxer) {
+    for (unsigned i = 0; i < demuxer->sink_count; ++i) {
+        struct sc_packet_sink *sink = demuxer->sinks[i];
+        if (sink->ops->disable) {
+            sink->ops->disable(sink);
+        }
+    }
+}
+
 static int
 run_demuxer(void *data) {
     struct sc_demuxer *demuxer = data;
 
-    enum AVCodecID codec_id = sc_demuxer_recv_codec_id(demuxer);
+    // Flag to report end-of-stream (i.e. device disconnected)
+    bool eos = false;
+
+    uint32_t raw_codec_id;
+    bool ok = sc_demuxer_recv_codec_id(demuxer, &raw_codec_id);
+    if (!ok) {
+        LOGE("Demuxer '%s': stream disabled due to connection error",
+             demuxer->name);
+        eos = true;
+        goto end;
+    }
+
+    if (raw_codec_id == 0) {
+        LOGW("Demuxer '%s': stream explicitly disabled by the device",
+             demuxer->name);
+        sc_demuxer_disable_sinks(demuxer);
+        eos = true;
+        goto end;
+    }
+
+    enum AVCodecID codec_id = sc_demuxer_to_avcodec_id(raw_codec_id);
     if (codec_id == AV_CODEC_ID_NONE) {
-        // Error already logged
+        LOGE("Demuxer '%s': stream disabled due to unsupported codec",
+             demuxer->name);
+        sc_demuxer_disable_sinks(demuxer);
         goto end;
     }
 
     const AVCodec *codec = avcodec_find_decoder(codec_id);
     if (!codec) {
-        LOGE("H.264 decoder not found");
-        goto end;
-    }
-
-    demuxer->codec_ctx = avcodec_alloc_context3(codec);
-    if (!demuxer->codec_ctx) {
-        LOG_OOM();
+        LOGE("Demuxer '%s': stream disabled due to missing decoder",
+             demuxer->name);
+        sc_demuxer_disable_sinks(demuxer);
         goto end;
     }
 
     if (!sc_demuxer_open_sinks(demuxer, codec)) {
-        LOGE("Could not open demuxer sinks");
-        goto finally_free_codec_ctx;
+        goto end;
     }
 
-    demuxer->parser = av_parser_init(codec_id);
-    if (!demuxer->parser) {
-        LOGE("Could not initialize parser");
-        goto finally_close_sinks;
-    }
+    // Config packets must be merged with the next non-config packet only for
+    // video streams
+    bool must_merge_config_packet = codec->type == AVMEDIA_TYPE_VIDEO;
 
-    // We must only pass complete frames to av_parser_parse2()!
-    // It's more complicated, but this allows to reduce the latency by 1 frame!
-    demuxer->parser->flags |= PARSER_FLAG_COMPLETE_FRAMES;
+    struct sc_packet_merger merger;
+
+    if (must_merge_config_packet) {
+        sc_packet_merger_init(&merger);
+    }
 
     AVPacket *packet = av_packet_alloc();
     if (!packet) {
         LOG_OOM();
-        goto finally_close_parser;
+        goto finally_close_sinks;
     }
 
     for (;;) {
         bool ok = sc_demuxer_recv_packet(demuxer, packet);
         if (!ok) {
             // end of stream
+            eos = true;
             break;
         }
 
+        if (must_merge_config_packet) {
+            // Prepend any config packet to the next media packet
+            ok = sc_packet_merger_merge(&merger, packet);
+            if (!ok) {
+                av_packet_unref(packet);
+                break;
+            }
+        }
+
         ok = sc_demuxer_push_packet(demuxer, packet);
         av_packet_unref(packet);
         if (!ok) {
@@ -250,35 +256,31 @@ run_demuxer(void *data) {
         }
     }
 
-    LOGD("End of frames");
+    LOGD("Demuxer '%s': end of frames", demuxer->name);
 
-    if (demuxer->pending) {
-        av_packet_free(&demuxer->pending);
+    if (must_merge_config_packet) {
+        sc_packet_merger_destroy(&merger);
     }
 
     av_packet_free(&packet);
-finally_close_parser:
-    av_parser_close(demuxer->parser);
 finally_close_sinks:
     sc_demuxer_close_sinks(demuxer);
-finally_free_codec_ctx:
-    avcodec_free_context(&demuxer->codec_ctx);
 end:
-    demuxer->cbs->on_eos(demuxer, demuxer->cbs_userdata);
+    demuxer->cbs->on_ended(demuxer, eos, demuxer->cbs_userdata);
 
     return 0;
 }
 
 void
-sc_demuxer_init(struct sc_demuxer *demuxer, sc_socket socket,
+sc_demuxer_init(struct sc_demuxer *demuxer, const char *name, sc_socket socket,
                 const struct sc_demuxer_callbacks *cbs, void *cbs_userdata) {
     assert(socket != SC_SOCKET_NONE);
 
+    demuxer->name = name; // statically allocated
     demuxer->socket = socket;
-    demuxer->pending = NULL;
     demuxer->sink_count = 0;
 
-    assert(cbs && cbs->on_eos);
+    assert(cbs && cbs->on_ended);
 
     demuxer->cbs = cbs;
     demuxer->cbs_userdata = cbs_userdata;
@@ -294,12 +296,12 @@ sc_demuxer_add_sink(struct sc_demuxer *demuxer, struct sc_packet_sink *sink) {
 
 bool
 sc_demuxer_start(struct sc_demuxer *demuxer) {
-    LOGD("Starting demuxer thread");
+    LOGD("Demuxer '%s': starting thread", demuxer->name);
 
     bool ok = sc_thread_create(&demuxer->thread, run_demuxer, "scrcpy-demuxer",
                                demuxer);
     if (!ok) {
-        LOGE("Could not start demuxer thread");
+        LOGE("Demuxer '%s': could not start thread", demuxer->name);
         return false;
     }
     return true;

app/src/demuxer.h

@@ -15,28 +15,25 @@
 #define SC_DEMUXER_MAX_SINKS 2
 
 struct sc_demuxer {
+    const char *name; // must be statically allocated (e.g. a string literal)
+
     sc_socket socket;
     sc_thread thread;
 
     struct sc_packet_sink *sinks[SC_DEMUXER_MAX_SINKS];
     unsigned sink_count;
 
-    AVCodecContext *codec_ctx;
-    AVCodecParserContext *parser;
-    // successive packets may need to be concatenated, until a non-config
-    // packet is available
-    AVPacket *pending;
-
     const struct sc_demuxer_callbacks *cbs;
     void *cbs_userdata;
 };
 
 struct sc_demuxer_callbacks {
-    void (*on_eos)(struct sc_demuxer *demuxer, void *userdata);
+    void (*on_ended)(struct sc_demuxer *demuxer, bool eos, void *userdata);
 };
 
+// The name must be statically allocated (e.g. a string literal)
 void
-sc_demuxer_init(struct sc_demuxer *demuxer, sc_socket socket,
+sc_demuxer_init(struct sc_demuxer *demuxer, const char *name, sc_socket socket,
                 const struct sc_demuxer_callbacks *cbs, void *cbs_userdata);
 
 void

app/src/events.h

@@ -1,5 +1,7 @@
-#define EVENT_NEW_FRAME                 SDL_USEREVENT
-#define EVENT_STREAM_STOPPED           (SDL_USEREVENT + 1)
-#define EVENT_SERVER_CONNECTION_FAILED (SDL_USEREVENT + 2)
-#define EVENT_SERVER_CONNECTED         (SDL_USEREVENT + 3)
-#define EVENT_USB_DEVICE_DISCONNECTED  (SDL_USEREVENT + 4)
+#define SC_EVENT_NEW_FRAME                 SDL_USEREVENT
+#define SC_EVENT_DEVICE_DISCONNECTED      (SDL_USEREVENT + 1)
+#define SC_EVENT_SERVER_CONNECTION_FAILED (SDL_USEREVENT + 2)
+#define SC_EVENT_SERVER_CONNECTED         (SDL_USEREVENT + 3)
+#define SC_EVENT_USB_DEVICE_DISCONNECTED  (SDL_USEREVENT + 4)
+#define SC_EVENT_DEMUXER_ERROR            (SDL_USEREVENT + 5)
+#define SC_EVENT_RECORDER_ERROR           (SDL_USEREVENT + 6)

app/src/options.c

@@ -14,6 +14,7 @@ const struct scrcpy_options scrcpy_options_default = {
 #endif
     .log_level = SC_LOG_LEVEL_INFO,
     .codec = SC_CODEC_H264,
+    .audio_codec = SC_CODEC_OPUS,
     .record_format = SC_RECORD_FORMAT_AUTO,
     .keyboard_input_mode = SC_KEYBOARD_INPUT_MODE_INJECT,
     .port_range = {
@@ -27,7 +28,8 @@ const struct scrcpy_options scrcpy_options_default = {
         .count = 2,
     },
     .max_size = 0,
-    .bit_rate = DEFAULT_BIT_RATE,
+    .bit_rate = 0,
+    .audio_bit_rate = 0,
     .max_fps = 0,
     .lock_video_orientation = SC_LOCK_VIDEO_ORIENTATION_UNLOCKED,
     .rotation = 0,

app/src/options.h

@@ -27,6 +27,8 @@ enum sc_codec {
     SC_CODEC_H264,
     SC_CODEC_H265,
     SC_CODEC_AV1,
+    SC_CODEC_OPUS,
+    SC_CODEC_AAC,
 };
 
 enum sc_lock_video_orientation {
@@ -100,6 +102,7 @@ struct scrcpy_options {
 #endif
     enum sc_log_level log_level;
     enum sc_codec codec;
+    enum sc_codec audio_codec;
     enum sc_record_format record_format;
     enum sc_keyboard_input_mode keyboard_input_mode;
     enum sc_mouse_input_mode mouse_input_mode;
@@ -109,6 +112,7 @@ struct scrcpy_options {
     struct sc_shortcut_mods shortcut_mods;
     uint16_t max_size;
     uint32_t bit_rate;
+    uint32_t audio_bit_rate;
     uint16_t max_fps;
     enum sc_lock_video_orientation lock_video_orientation;
     uint8_t rotation;

app/src/packet_merger.c

@@ -0,0 +1,48 @@
+#include "packet_merger.h"
+
+#include "util/log.h"
+
+void
+sc_packet_merger_init(struct sc_packet_merger *merger) {
+    merger->config = NULL;
+}
+
+void
+sc_packet_merger_destroy(struct sc_packet_merger *merger) {
+    free(merger->config);
+}
+
+bool
+sc_packet_merger_merge(struct sc_packet_merger *merger, AVPacket *packet) {
+    bool is_config = packet->pts == AV_NOPTS_VALUE;
+
+    if (is_config) {
+        free(merger->config);
+
+        merger->config = malloc(packet->size);
+        if (!merger->config) {
+            LOG_OOM();
+            return false;
+        }
+
+        memcpy(merger->config, packet->data, packet->size);
+        merger->config_size = packet->size;
+    } else if (merger->config) {
+        size_t config_size = merger->config_size;
+        size_t media_size = packet->size;
+
+        if (av_grow_packet(packet, config_size)) {
+            LOG_OOM();
+            return false;
+        }
+
+        memmove(packet->data + config_size, packet->data, media_size);
+        memcpy(packet->data, merger->config, config_size);
+
+        free(merger->config);
+        merger->config = NULL;
+        // merger->size is meaningless when merger->config is NULL
+    }
+
+    return true;
+}

app/src/packet_merger.h

@@ -0,0 +1,43 @@
+#ifndef SC_PACKET_MERGER_H
+#define SC_PACKET_MERGER_H
+
+#include "common.h"
+
+#include <stdbool.h>
+#include <stdint.h>
+#include <libavcodec/avcodec.h>
+
+/**
+ * Config packets (containing the SPS/PPS) are sent in-band. A new config
+ * packet is sent whenever a new encoding session is started (on start and on
+ * device orientation change).
+ *
+ * Every time a config packet is received, it must be sent alone (for recorder
+ * extradata), then concatenated to the next media packet (for correct decoding
+ * and recording).
+ *
+ * This helper reads every input packet and modifies each media packet which
+ * immediately follows a config packet to prepend the config packet payload.
+ */
+
+struct sc_packet_merger {
+    uint8_t *config;
+    size_t config_size;
+};
+
+void
+sc_packet_merger_init(struct sc_packet_merger *merger);
+
+void
+sc_packet_merger_destroy(struct sc_packet_merger *merger);
+
+/**
+ * If the packet is a config packet, then keep its data for later.
+ * Otherwise (if the packet is a media packet), then if a config packet is
+ * pending, prepend the config packet to this packet (so the packet is
+ * modified!).
+ */
+bool
+sc_packet_merger_merge(struct sc_packet_merger *merger, AVPacket *packet);
+
+#endif

app/src/receiver.c

@@ -7,7 +7,7 @@
 #include "util/log.h"
 
 bool
-receiver_init(struct receiver *receiver, sc_socket control_socket,
+sc_receiver_init(struct sc_receiver *receiver, sc_socket control_socket,
               struct sc_acksync *acksync) {
     bool ok = sc_mutex_init(&receiver->mutex);
     if (!ok) {
@@ -21,12 +21,12 @@ receiver_init(struct receiver *receiver, sc_socket control_socket,
 }
 
 void
-receiver_destroy(struct receiver *receiver) {
+sc_receiver_destroy(struct sc_receiver *receiver) {
     sc_mutex_destroy(&receiver->mutex);
 }
 
 static void
-process_msg(struct receiver *receiver, struct device_msg *msg) {
+process_msg(struct sc_receiver *receiver, struct device_msg *msg) {
     switch (msg->type) {
         case DEVICE_MSG_TYPE_CLIPBOARD: {
             char *current = SDL_GetClipboardText();
@@ -51,7 +51,7 @@ process_msg(struct receiver *receiver, struct device_msg *msg) {
 }
 
 static ssize_t
-process_msgs(struct receiver *receiver, const unsigned char *buf, size_t len) {
+process_msgs(struct sc_receiver *receiver, const unsigned char *buf, size_t len) {
     size_t head = 0;
     for (;;) {
         struct device_msg msg;
@@ -76,7 +76,7 @@ process_msgs(struct receiver *receiver, const unsigned char *buf, size_t len) {
 
 static int
 run_receiver(void *data) {
-    struct receiver *receiver = data;
+    struct sc_receiver *receiver = data;
 
     static unsigned char buf[DEVICE_MSG_MAX_SIZE];
     size_t head = 0;
@@ -108,7 +108,7 @@ run_receiver(void *data) {
 }
 
 bool
-receiver_start(struct receiver *receiver) {
+sc_receiver_start(struct sc_receiver *receiver) {
     LOGD("Starting receiver thread");
 
     bool ok = sc_thread_create(&receiver->thread, run_receiver,
@@ -122,6 +122,6 @@ receiver_start(struct receiver *receiver) {
 }
 
 void
-receiver_join(struct receiver *receiver) {
+sc_receiver_join(struct sc_receiver *receiver) {
     sc_thread_join(&receiver->thread, NULL);
 }

app/src/receiver.h

@@ -11,7 +11,7 @@
 
 // receive events from the device
 // managed by the controller
-struct receiver {
+struct sc_receiver {
     sc_socket control_socket;
     sc_thread thread;
     sc_mutex mutex;
@@ -20,18 +20,18 @@ struct receiver {
 };
 
 bool
-receiver_init(struct receiver *receiver, sc_socket control_socket,
-              struct sc_acksync *acksync);
+sc_receiver_init(struct sc_receiver *receiver, sc_socket control_socket,
+                 struct sc_acksync *acksync);
 
 void
-receiver_destroy(struct receiver *receiver);
+sc_receiver_destroy(struct sc_receiver *receiver);
 
 bool
-receiver_start(struct receiver *receiver);
+sc_receiver_start(struct sc_receiver *receiver);
 
-// no receiver_stop(), it will automatically stop on control_socket shutdown
+// no sc_receiver_stop(), it will automatically stop on control_socket shutdown
 
 void
-receiver_join(struct receiver *receiver);
+sc_receiver_join(struct sc_receiver *receiver);
 
 #endif

app/src/recorder.c

@@ -8,10 +8,11 @@
 #include "util/log.h"
 #include "util/str.h"
 
-/** Downcast packet_sink to recorder */
-#define DOWNCAST(SINK) container_of(SINK, struct sc_recorder, packet_sink)
-
-#define SC_PTS_ORIGIN_NONE UINT64_C(-1)
+/** Downcast packet sinks to recorder */
+#define DOWNCAST_VIDEO(SINK) \
+    container_of(SINK, struct sc_recorder, video_packet_sink)
+#define DOWNCAST_AUDIO(SINK) \
+    container_of(SINK, struct sc_recorder, audio_packet_sink)
 
 static const AVRational SCRCPY_TIME_BASE = {1, 1000000}; // timestamps in us
 
@@ -80,9 +81,7 @@ sc_recorder_get_format_name(enum sc_record_format format) {
 }
 
 static bool
-sc_recorder_write_header(struct sc_recorder *recorder, const AVPacket *packet) {
-    AVStream *ostream = recorder->ctx->streams[0];
-
+sc_recorder_set_extradata(AVStream *ostream, const AVPacket *packet) {
     uint8_t *extradata = av_malloc(packet->size * sizeof(uint8_t));
     if (!extradata) {
         LOG_OOM();
@@ -94,183 +93,56 @@ sc_recorder_write_header(struct sc_recorder *recorder, const AVPacket *packet) {
 
     ostream->codecpar->extradata = extradata;
     ostream->codecpar->extradata_size = packet->size;
-
-    int ret = avformat_write_header(recorder->ctx, NULL);
-    if (ret < 0) {
-        LOGE("Failed to write header to %s", recorder->filename);
-        return false;
-    }
-
     return true;
 }
 
-static void
-sc_recorder_rescale_packet(struct sc_recorder *recorder, AVPacket *packet) {
-    AVStream *ostream = recorder->ctx->streams[0];
-    av_packet_rescale_ts(packet, SCRCPY_TIME_BASE, ostream->time_base);
+static inline void
+sc_recorder_rescale_packet(AVStream *stream, AVPacket *packet) {
+    av_packet_rescale_ts(packet, SCRCPY_TIME_BASE, stream->time_base);
 }
 
 static bool
-sc_recorder_write(struct sc_recorder *recorder, AVPacket *packet) {
-    if (!recorder->header_written) {
-        if (packet->pts != AV_NOPTS_VALUE) {
-            LOGE("The first packet is not a config packet");
-            return false;
-        }
-        bool ok = sc_recorder_write_header(recorder, packet);
-        if (!ok) {
-            return false;
-        }
-        recorder->header_written = true;
-        return true;
-    }
-
-    if (packet->pts == AV_NOPTS_VALUE) {
-        // ignore config packets
-        return true;
-    }
-
-    sc_recorder_rescale_packet(recorder, packet);
-    return av_write_frame(recorder->ctx, packet) >= 0;
+sc_recorder_write_stream(struct sc_recorder *recorder, int stream_index,
+                         AVPacket *packet) {
+    AVStream *stream = recorder->ctx->streams[stream_index];
+    sc_recorder_rescale_packet(stream, packet);
+    return av_interleaved_write_frame(recorder->ctx, packet) >= 0;
 }
 
-static int
-run_recorder(void *data) {
-    struct sc_recorder *recorder = data;
+static inline bool
+sc_recorder_write_video(struct sc_recorder *recorder, AVPacket *packet) {
+    return sc_recorder_write_stream(recorder, recorder->video_stream_index,
+                                    packet);
+}
 
-    for (;;) {
-        sc_mutex_lock(&recorder->mutex);
-
-        while (!recorder->stopped && sc_queue_is_empty(&recorder->queue)) {
-            sc_cond_wait(&recorder->queue_cond, &recorder->mutex);
-        }
-
-        // if stopped is set, continue to process the remaining events (to
-        // finish the recording) before actually stopping
-
-        if (recorder->stopped && sc_queue_is_empty(&recorder->queue)) {
-            sc_mutex_unlock(&recorder->mutex);
-            struct sc_record_packet *last = recorder->previous;
-            if (last) {
-                // assign an arbitrary duration to the last packet
-                last->packet->duration = 100000;
-                bool ok = sc_recorder_write(recorder, last->packet);
-                if (!ok) {
-                    // failing to write the last frame is not very serious, no
-                    // future frame may depend on it, so the resulting file
-                    // will still be valid
-                    LOGW("Could not record last packet");
-                }
-                sc_record_packet_delete(last);
-            }
-            break;
-        }
-
-        struct sc_record_packet *rec;
-        sc_queue_take(&recorder->queue, next, &rec);
-
-        sc_mutex_unlock(&recorder->mutex);
-
-        if (recorder->pts_origin == SC_PTS_ORIGIN_NONE
-                && rec->packet->pts != AV_NOPTS_VALUE) {
-            // First PTS received
-            recorder->pts_origin = rec->packet->pts;
-        }
-
-        if (rec->packet->pts != AV_NOPTS_VALUE) {
-            // Set PTS relatve to the origin
-            rec->packet->pts -= recorder->pts_origin;
-            rec->packet->dts = rec->packet->pts;
-        }
-
-        // recorder->previous is only written from this thread, no need to lock
-        struct sc_record_packet *previous = recorder->previous;
-        recorder->previous = rec;
-
-        if (!previous) {
-            // we just received the first packet
-            continue;
-        }
-
-        // config packets have no PTS, we must ignore them
-        if (rec->packet->pts != AV_NOPTS_VALUE
-            && previous->packet->pts != AV_NOPTS_VALUE) {
-            // we now know the duration of the previous packet
-            previous->packet->duration =
-                rec->packet->pts - previous->packet->pts;
-        }
-
-        bool ok = sc_recorder_write(recorder, previous->packet);
-        sc_record_packet_delete(previous);
-        if (!ok) {
-            LOGE("Could not record packet");
-
-            sc_mutex_lock(&recorder->mutex);
-            recorder->failed = true;
-            // discard pending packets
-            sc_recorder_queue_clear(&recorder->queue);
-            sc_mutex_unlock(&recorder->mutex);
-            break;
-        }
-    }
-
-    if (!recorder->failed) {
-        if (recorder->header_written) {
-            int ret = av_write_trailer(recorder->ctx);
-            if (ret < 0) {
-                LOGE("Failed to write trailer to %s", recorder->filename);
-                recorder->failed = true;
-            }
-        } else {
-            // the recorded file is empty
-            recorder->failed = true;
-        }
-    }
-
-    if (recorder->failed) {
-        LOGE("Recording failed to %s", recorder->filename);
-    } else {
-        const char *format_name = sc_recorder_get_format_name(recorder->format);
-        LOGI("Recording complete to %s file: %s", format_name,
-                                                  recorder->filename);
-    }
-
-    LOGD("Recorder thread ended");
-
-    return 0;
+static inline bool
+sc_recorder_write_audio(struct sc_recorder *recorder, AVPacket *packet) {
+    return sc_recorder_write_stream(recorder, recorder->audio_stream_index,
+                                    packet);
 }
 
 static bool
-sc_recorder_open(struct sc_recorder *recorder, const AVCodec *input_codec) {
-    bool ok = sc_mutex_init(&recorder->mutex);
-    if (!ok) {
-        return false;
-    }
-
-    ok = sc_cond_init(&recorder->queue_cond);
-    if (!ok) {
-        goto error_mutex_destroy;
-    }
-
-    sc_queue_init(&recorder->queue);
-    recorder->stopped = false;
-    recorder->failed = false;
-    recorder->header_written = false;
-    recorder->previous = NULL;
-    recorder->pts_origin = SC_PTS_ORIGIN_NONE;
-
+sc_recorder_open_output_file(struct sc_recorder *recorder) {
     const char *format_name = sc_recorder_get_format_name(recorder->format);
     assert(format_name);
     const AVOutputFormat *format = find_muxer(format_name);
     if (!format) {
         LOGE("Could not find muxer");
-        goto error_cond_destroy;
+        return false;
     }
 
     recorder->ctx = avformat_alloc_context();
     if (!recorder->ctx) {
         LOG_OOM();
-        goto error_cond_destroy;
+        return false;
+    }
+
+    int ret = avio_open(&recorder->ctx->pb, recorder->filename,
+                        AVIO_FLAG_WRITE);
+    if (ret < 0) {
+        LOGE("Failed to open output file: %s", recorder->filename);
+        avformat_free_context(recorder->ctx);
+        return false;
     }
 
     // contrary to the deprecated API (av_oformat_next()), av_muxer_iterate()
@@ -282,71 +154,436 @@ sc_recorder_open(struct sc_recorder *recorder, const AVCodec *input_codec) {
     av_dict_set(&recorder->ctx->metadata, "comment",
                 "Recorded by scrcpy " SCRCPY_VERSION, 0);
 
-    AVStream *ostream = avformat_new_stream(recorder->ctx, input_codec);
-    if (!ostream) {
-        goto error_avformat_free_context;
-    }
-
-    ostream->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
-    ostream->codecpar->codec_id = input_codec->id;
-    ostream->codecpar->format = AV_PIX_FMT_YUV420P;
-    ostream->codecpar->width = recorder->declared_frame_size.width;
-    ostream->codecpar->height = recorder->declared_frame_size.height;
-
-    int ret = avio_open(&recorder->ctx->pb, recorder->filename,
-                        AVIO_FLAG_WRITE);
-    if (ret < 0) {
-        LOGE("Failed to open output file: %s", recorder->filename);
-        // ostream will be cleaned up during context cleaning
-        goto error_avformat_free_context;
-    }
-
-    LOGD("Starting recorder thread");
-    ok = sc_thread_create(&recorder->thread, run_recorder, "scrcpy-recorder",
-                          recorder);
-    if (!ok) {
-        LOGE("Could not start recorder thread");
-        goto error_avio_close;
-    }
-
     LOGI("Recording started to %s file: %s", format_name, recorder->filename);
-
     return true;
-
-error_avio_close:
-    avio_close(recorder->ctx->pb);
-error_avformat_free_context:
-    avformat_free_context(recorder->ctx);
-error_cond_destroy:
-    sc_cond_destroy(&recorder->queue_cond);
-error_mutex_destroy:
-    sc_mutex_destroy(&recorder->mutex);
-
-    return false;
 }
 
 static void
-sc_recorder_close(struct sc_recorder *recorder) {
-    sc_mutex_lock(&recorder->mutex);
-    recorder->stopped = true;
-    sc_cond_signal(&recorder->queue_cond);
-    sc_mutex_unlock(&recorder->mutex);
-
-    sc_thread_join(&recorder->thread, NULL);
-
+sc_recorder_close_output_file(struct sc_recorder *recorder) {
     avio_close(recorder->ctx->pb);
     avformat_free_context(recorder->ctx);
-    sc_cond_destroy(&recorder->queue_cond);
-    sc_mutex_destroy(&recorder->mutex);
 }
 
 static bool
-sc_recorder_push(struct sc_recorder *recorder, const AVPacket *packet) {
+sc_recorder_wait_video_stream(struct sc_recorder *recorder) {
     sc_mutex_lock(&recorder->mutex);
-    assert(!recorder->stopped);
+    while (!recorder->video_codec && !recorder->stopped) {
+        sc_cond_wait(&recorder->stream_cond, &recorder->mutex);
+    }
+    const AVCodec *codec = recorder->video_codec;
+    sc_mutex_unlock(&recorder->mutex);
 
-    if (recorder->failed) {
-        // reject any new packet (this will stop the stream)
+    if (codec) {
+        AVStream *stream = avformat_new_stream(recorder->ctx, codec);
+        if (!stream) {
+            return false;
+        }
+
+        stream->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
+        stream->codecpar->codec_id = codec->id;
+        stream->codecpar->format = AV_PIX_FMT_YUV420P;
+        stream->codecpar->width = recorder->declared_frame_size.width;
+        stream->codecpar->height = recorder->declared_frame_size.height;
+
+        recorder->video_stream_index = stream->index;
+    }
+
+    return true;
+}
+
+static bool
+sc_recorder_wait_audio_stream(struct sc_recorder *recorder) {
+    sc_mutex_lock(&recorder->mutex);
+    while (!recorder->audio_codec && !recorder->audio_disabled
+            && !recorder->stopped) {
+        sc_cond_wait(&recorder->stream_cond, &recorder->mutex);
+    }
+
+    if (recorder->audio_disabled) {
+        // Reset audio flag. From there, the recorder thread may access this
+        // flag without any mutex.
+        recorder->audio = false;
+    }
+
+    const AVCodec *codec = recorder->audio_codec;
+    sc_mutex_unlock(&recorder->mutex);
+
+    if (codec) {
+        AVStream *stream = avformat_new_stream(recorder->ctx, codec);
+        if (!stream) {
+            return false;
+        }
+
+        stream->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
+        stream->codecpar->codec_id = codec->id;
+        stream->codecpar->ch_layout.nb_channels = 2;
+        stream->codecpar->sample_rate = 48000;
+
+        recorder->audio_stream_index = stream->index;
+    }
+
+    return true;
+}
+
+static inline bool
+sc_recorder_has_empty_queues(struct sc_recorder *recorder) {
+    if (sc_queue_is_empty(&recorder->video_queue)) {
+        // The video queue is empty
+        return true;
+    }
+
+    if (recorder->audio && sc_queue_is_empty(&recorder->audio_queue)) {
+        // The audio queue is empty (when audio is enabled)
+        return true;
+    }
+
+    // No queue is empty
+    return false;
+}
+
+static bool
+sc_recorder_process_header(struct sc_recorder *recorder) {
+    sc_mutex_lock(&recorder->mutex);
+
+    while (!recorder->stopped && sc_recorder_has_empty_queues(recorder)) {
+        sc_cond_wait(&recorder->queue_cond, &recorder->mutex);
+    }
+
+    if (recorder->stopped && sc_queue_is_empty(&recorder->video_queue)) {
+        // If the recorder is stopped, don't process anything if there are not
+        // at least video packets
+        sc_mutex_unlock(&recorder->mutex);
+        return false;
+    }
+
+    struct sc_record_packet *video_pkt;
+    sc_queue_take(&recorder->video_queue, next, &video_pkt);
+
+    struct sc_record_packet *audio_pkt = NULL;
+    if (!sc_queue_is_empty(&recorder->audio_queue)) {
+        assert(recorder->audio);
+        sc_queue_take(&recorder->audio_queue, next, &audio_pkt);
+    }
+
+    sc_mutex_unlock(&recorder->mutex);
+
+    int ret = false;
+
+    if (video_pkt->packet->pts != AV_NOPTS_VALUE) {
+        LOGE("The first video packet is not a config packet");
+        goto end;
+    }
+
+    assert(recorder->video_stream_index >= 0);
+    AVStream *video_stream =
+        recorder->ctx->streams[recorder->video_stream_index];
+    bool ok = sc_recorder_set_extradata(video_stream, video_pkt->packet);
+    if (!ok) {
+        goto end;
+    }
+
+    if (audio_pkt) {
+        if (audio_pkt->packet->pts != AV_NOPTS_VALUE) {
+            LOGE("The first audio packet is not a config packet");
+            goto end;
+        }
+
+        assert(recorder->audio_stream_index >= 0);
+        AVStream *audio_stream =
+            recorder->ctx->streams[recorder->audio_stream_index];
+        ok = sc_recorder_set_extradata(audio_stream, audio_pkt->packet);
+        if (!ok) {
+            goto end;
+        }
+    }
+
+    ok = avformat_write_header(recorder->ctx, NULL) >= 0;
+    if (!ok) {
+        LOGE("Failed to write header to %s", recorder->filename);
+        goto end;
+    }
+
+    ret = true;
+
+end:
+    sc_record_packet_delete(video_pkt);
+    if (audio_pkt) {
+        sc_record_packet_delete(audio_pkt);
+    }
+
+    return ret;
+}
+
+static bool
+sc_recorder_process_packets(struct sc_recorder *recorder) {
+    int64_t pts_origin = AV_NOPTS_VALUE;
+
+    bool header_written = sc_recorder_process_header(recorder);
+    if (!header_written) {
+        return false;
+    }
+
+    struct sc_record_packet *video_pkt = NULL;
+    struct sc_record_packet *audio_pkt = NULL;
+
+    // We can write a video packet only once we received the next one so that
+    // we can set its duration (next_pts - current_pts)
+    struct sc_record_packet *video_pkt_previous = NULL;
+
+    bool error = false;
+
+    for (;;) {
+        sc_mutex_lock(&recorder->mutex);
+
+        while (!recorder->stopped) {
+            if (!video_pkt && !sc_queue_is_empty(&recorder->video_queue)) {
+                // A new packet may be assigned to video_pkt and be processed
+                break;
+            }
+            if (recorder->audio && !audio_pkt
+                    && !sc_queue_is_empty(&recorder->audio_queue)) {
+                // A new packet may be assigned to audio_pkt and be processed
+                break;
+            }
+            sc_cond_wait(&recorder->queue_cond, &recorder->mutex);
+        }
+
+        // If stopped is set, continue to process the remaining events (to
+        // finish the recording) before actually stopping.
+
+        // If there is no audio, then the audio_queue will remain empty forever
+        // and audio_pkt will always be NULL.
+        assert(recorder->audio
+                || (!audio_pkt && sc_queue_is_empty(&recorder->audio_queue)));
+
+        if (!video_pkt && !sc_queue_is_empty(&recorder->video_queue)) {
+            sc_queue_take(&recorder->video_queue, next, &video_pkt);
+        }
+
+        if (!audio_pkt && !sc_queue_is_empty(&recorder->audio_queue)) {
+            sc_queue_take(&recorder->audio_queue, next, &audio_pkt);
+        }
+
+        if (recorder->stopped && !video_pkt && !audio_pkt) {
+            assert(sc_queue_is_empty(&recorder->video_queue));
+            assert(sc_queue_is_empty(&recorder->audio_queue));
+            sc_mutex_unlock(&recorder->mutex);
+            break;
+        }
+
+        assert(video_pkt || audio_pkt); // at least one
+
+        sc_mutex_unlock(&recorder->mutex);
+
+        // Ignore further config packets (e.g. on device orientation
+        // change). The next non-config packet will have the config packet
+        // data prepended.
+        if (video_pkt && video_pkt->packet->pts == AV_NOPTS_VALUE) {
+            sc_record_packet_delete(video_pkt);
+            video_pkt = NULL;
+        }
+
+        if (audio_pkt && audio_pkt->packet->pts == AV_NOPTS_VALUE) {
+            sc_record_packet_delete(audio_pkt);
+            audio_pkt= NULL;
+        }
+
+        if (pts_origin == AV_NOPTS_VALUE) {
+            if (!recorder->audio) {
+                assert(video_pkt);
+                pts_origin = video_pkt->packet->pts;
+            } else if (video_pkt && audio_pkt) {
+                pts_origin =
+                    MIN(video_pkt->packet->pts, audio_pkt->packet->pts);
+            } else if (recorder->stopped) {
+                if (video_pkt) {
+                    // The recorder is stopped without audio, record the video
+                    // packets
+                    pts_origin = video_pkt->packet->pts;
+                } else {
+                    // Fail if there is no video
+                    error = true;
+                    goto end;
+                }
+                // If the recorder is stopped while one of the streams has no
+                // packets, then we must avoid a live-loop and correctly record
+                // the stream having packets.
+                pts_origin = video_pkt ? video_pkt->packet->pts
+                                       : audio_pkt->packet->pts;
+            } else {
+                // We need both video and audio packets to initialize pts_origin
+                continue;
+            }
+        }
+
+        assert(pts_origin != AV_NOPTS_VALUE);
+
+        if (video_pkt) {
+            video_pkt->packet->pts -= pts_origin;
+            video_pkt->packet->dts = video_pkt->packet->pts;
+
+            if (video_pkt_previous) {
+                // we now know the duration of the previous packet
+                video_pkt_previous->packet->duration =
+                    video_pkt->packet->pts - video_pkt_previous->packet->pts;
+
+                bool ok = sc_recorder_write_video(recorder,
+                                                  video_pkt_previous->packet);
+                sc_record_packet_delete(video_pkt_previous);
+                if (!ok) {
+                    LOGE("Could not record video packet");
+                    error = true;
+                    goto end;
+                }
+            }
+
+            video_pkt_previous = video_pkt;
+            video_pkt = NULL;
+        }
+
+        if (audio_pkt) {
+            audio_pkt->packet->pts -= pts_origin;
+            audio_pkt->packet->dts = audio_pkt->packet->pts;
+
+            bool ok = sc_recorder_write_audio(recorder, audio_pkt->packet);
+            if (!ok) {
+                LOGE("Could not record audio packet");
+                error = true;
+                goto end;
+            }
+
+            sc_record_packet_delete(audio_pkt);
+            audio_pkt = NULL;
+        }
+    }
+
+    // Write the last video packet
+    struct sc_record_packet *last = video_pkt_previous;
+    if (last) {
+        // assign an arbitrary duration to the last packet
+        last->packet->duration = 100000;
+        bool ok = sc_recorder_write_video(recorder, last->packet);
+        if (!ok) {
+            // failing to write the last frame is not very serious, no
+            // future frame may depend on it, so the resulting file
+            // will still be valid
+            LOGW("Could not record last packet");
+        }
+        sc_record_packet_delete(last);
+    }
+
+    int ret = av_write_trailer(recorder->ctx);
+    if (ret < 0) {
+        LOGE("Failed to write trailer to %s", recorder->filename);
+        error = false;
+    }
+
+end:
+    if (video_pkt) {
+        sc_record_packet_delete(video_pkt);
+    }
+    if (audio_pkt) {
+        sc_record_packet_delete(audio_pkt);
+    }
+
+    return !error;
+}
+
+static bool
+sc_recorder_record(struct sc_recorder *recorder) {
+    bool ok = sc_recorder_open_output_file(recorder);
+    if (!ok) {
+        return false;
+    }
+
+    ok = sc_recorder_wait_video_stream(recorder);
+    if (!ok) {
+        sc_recorder_close_output_file(recorder);
+        return false;
+    }
+
+    if (recorder->audio) {
+        ok = sc_recorder_wait_audio_stream(recorder);
+        if (!ok) {
+            sc_recorder_close_output_file(recorder);
+            return false;
+        }
+    }
+
+    // If recorder->stopped, process any queued packet anyway
+
+    ok = sc_recorder_process_packets(recorder);
+    sc_recorder_close_output_file(recorder);
+    return ok;
+}
+
+static int
+run_recorder(void *data) {
+    struct sc_recorder *recorder = data;
+
+    bool success = sc_recorder_record(recorder);
+
+    sc_mutex_lock(&recorder->mutex);
+    // Prevent the producer to push any new packet
+    recorder->stopped = true;
+    // Discard pending packets
+    sc_recorder_queue_clear(&recorder->video_queue);
+    sc_mutex_unlock(&recorder->mutex);
+
+    if (success) {
+        const char *format_name = sc_recorder_get_format_name(recorder->format);
+        LOGI("Recording complete to %s file: %s", format_name,
+                                                  recorder->filename);
+    } else {
+        LOGE("Recording failed to %s", recorder->filename);
+    }
+
+    LOGD("Recorder thread ended");
+
+    recorder->cbs->on_ended(recorder, success, recorder->cbs_userdata);
+
+    return 0;
+}
+
+static bool
+sc_recorder_video_packet_sink_open(struct sc_packet_sink *sink,
+                                   const AVCodec *codec) {
+    struct sc_recorder *recorder = DOWNCAST_VIDEO(sink);
+    assert(codec);
+
+    sc_mutex_lock(&recorder->mutex);
+    if (recorder->stopped) {
+        sc_mutex_unlock(&recorder->mutex);
+        return false;
+    }
+
+    recorder->video_codec = codec;
+    sc_cond_signal(&recorder->stream_cond);
+    sc_mutex_unlock(&recorder->mutex);
+
+    return true;
+}
+
+static void
+sc_recorder_video_packet_sink_close(struct sc_packet_sink *sink) {
+    struct sc_recorder *recorder = DOWNCAST_VIDEO(sink);
+
+    sc_mutex_lock(&recorder->mutex);
+    // EOS also stops the recorder
+    recorder->stopped = true;
+    sc_cond_signal(&recorder->queue_cond);
+    sc_mutex_unlock(&recorder->mutex);
+}
+
+static bool
+sc_recorder_video_packet_sink_push(struct sc_packet_sink *sink,
+                                   const AVPacket *packet) {
+    struct sc_recorder *recorder = DOWNCAST_VIDEO(sink);
+
+    sc_mutex_lock(&recorder->mutex);
+
+    if (recorder->stopped) {
+        // reject any new packet
         sc_mutex_unlock(&recorder->mutex);
         return false;
     }
@@ -358,7 +595,9 @@ sc_recorder_push(struct sc_recorder *recorder, const AVPacket *packet) {
         return false;
     }
 
-    sc_queue_push(&recorder->queue, next, rec);
+    rec->packet->stream_index = 0;
+
+    sc_queue_push(&recorder->video_queue, next, rec);
     sc_cond_signal(&recorder->queue_cond);
 
     sc_mutex_unlock(&recorder->mutex);
@@ -366,51 +605,190 @@ sc_recorder_push(struct sc_recorder *recorder, const AVPacket *packet) {
 }
 
 static bool
-sc_recorder_packet_sink_open(struct sc_packet_sink *sink,
-                             const AVCodec *codec) {
-    struct sc_recorder *recorder = DOWNCAST(sink);
-    return sc_recorder_open(recorder, codec);
+sc_recorder_audio_packet_sink_open(struct sc_packet_sink *sink,
+                                   const AVCodec *codec) {
+    struct sc_recorder *recorder = DOWNCAST_AUDIO(sink);
+    assert(recorder->audio);
+    // only written from this thread, no need to lock
+    assert(!recorder->audio_disabled);
+    assert(codec);
+
+    sc_mutex_lock(&recorder->mutex);
+    recorder->audio_codec = codec;
+    sc_cond_signal(&recorder->stream_cond);
+    sc_mutex_unlock(&recorder->mutex);
+
+    return true;
 }
 
 static void
-sc_recorder_packet_sink_close(struct sc_packet_sink *sink) {
-    struct sc_recorder *recorder = DOWNCAST(sink);
-    sc_recorder_close(recorder);
+sc_recorder_audio_packet_sink_close(struct sc_packet_sink *sink) {
+    struct sc_recorder *recorder = DOWNCAST_AUDIO(sink);
+    assert(recorder->audio);
+    // only written from this thread, no need to lock
+    assert(!recorder->audio_disabled);
+
+    sc_mutex_lock(&recorder->mutex);
+    // EOS also stops the recorder
+    // TODO must stop only once both video and audio stream are complete
+    recorder->stopped = true;
+    sc_cond_signal(&recorder->queue_cond);
+    sc_mutex_unlock(&recorder->mutex);
 }
 
 static bool
-sc_recorder_packet_sink_push(struct sc_packet_sink *sink,
-                             const AVPacket *packet) {
-    struct sc_recorder *recorder = DOWNCAST(sink);
-    return sc_recorder_push(recorder, packet);
+sc_recorder_audio_packet_sink_push(struct sc_packet_sink *sink,
+                                   const AVPacket *packet) {
+    struct sc_recorder *recorder = DOWNCAST_AUDIO(sink);
+    assert(recorder->audio);
+    // only written from this thread, no need to lock
+    assert(!recorder->audio_disabled);
+
+    sc_mutex_lock(&recorder->mutex);
+
+    if (recorder->stopped) {
+        // reject any new packet
+        sc_mutex_unlock(&recorder->mutex);
+        return false;
+    }
+
+    struct sc_record_packet *rec = sc_record_packet_new(packet);
+    if (!rec) {
+        LOG_OOM();
+        sc_mutex_unlock(&recorder->mutex);
+        return false;
+    }
+
+    rec->packet->stream_index = 1;
+
+    sc_queue_push(&recorder->audio_queue, next, rec);
+    sc_cond_signal(&recorder->queue_cond);
+
+    sc_mutex_unlock(&recorder->mutex);
+    return true;
+}
+
+static void
+sc_recorder_audio_packet_sink_disable(struct sc_packet_sink *sink) {
+    struct sc_recorder *recorder = DOWNCAST_AUDIO(sink);
+    assert(recorder->audio);
+    // only written from this thread, no need to lock
+    assert(!recorder->audio_disabled);
+    assert(!recorder->audio_codec);
+
+    LOGW("Audio stream recording disabled");
+
+    sc_mutex_lock(&recorder->mutex);
+    recorder->audio_disabled = true;
+    sc_cond_signal(&recorder->stream_cond);
+    sc_mutex_unlock(&recorder->mutex);
 }
 
 bool
-sc_recorder_init(struct sc_recorder *recorder,
-                 const char *filename,
-                 enum sc_record_format format,
-                 struct sc_size declared_frame_size) {
+sc_recorder_init(struct sc_recorder *recorder, const char *filename,
+                 enum sc_record_format format, bool audio,
+                 struct sc_size declared_frame_size,
+                 const struct sc_recorder_callbacks *cbs, void *cbs_userdata) {
     recorder->filename = strdup(filename);
     if (!recorder->filename) {
         LOG_OOM();
         return false;
     }
 
+    bool ok = sc_mutex_init(&recorder->mutex);
+    if (!ok) {
+        goto error_free_filename;
+    }
+
+    ok = sc_cond_init(&recorder->queue_cond);
+    if (!ok) {
+        goto error_mutex_destroy;
+    }
+
+    ok = sc_cond_init(&recorder->stream_cond);
+    if (!ok) {
+        goto error_queue_cond_destroy;
+    }
+
+    recorder->audio = audio;
+
+    sc_queue_init(&recorder->video_queue);
+    sc_queue_init(&recorder->audio_queue);
+    recorder->stopped = false;
+
+    recorder->video_codec = NULL;
+    recorder->audio_codec = NULL;
+    recorder->audio_disabled = false;
+
+    recorder->video_stream_index = -1;
+    recorder->audio_stream_index = -1;
+
     recorder->format = format;
     recorder->declared_frame_size = declared_frame_size;
 
-    static const struct sc_packet_sink_ops ops = {
-        .open = sc_recorder_packet_sink_open,
-        .close = sc_recorder_packet_sink_close,
-        .push = sc_recorder_packet_sink_push,
+    assert(cbs && cbs->on_ended);
+
+    recorder->cbs = cbs;
+    recorder->cbs_userdata = cbs_userdata;
+
+    static const struct sc_packet_sink_ops video_ops = {
+        .open = sc_recorder_video_packet_sink_open,
+        .close = sc_recorder_video_packet_sink_close,
+        .push = sc_recorder_video_packet_sink_push,
     };
 
-    recorder->packet_sink.ops = &ops;
+    recorder->video_packet_sink.ops = &video_ops;
+
+    if (audio) {
+        static const struct sc_packet_sink_ops audio_ops = {
+            .open = sc_recorder_audio_packet_sink_open,
+            .close = sc_recorder_audio_packet_sink_close,
+            .push = sc_recorder_audio_packet_sink_push,
+            .disable = sc_recorder_audio_packet_sink_disable,
+        };
+
+        recorder->audio_packet_sink.ops = &audio_ops;
+    }
+
+    ok = sc_thread_create(&recorder->thread, run_recorder, "scrcpy-recorder",
+                          recorder);
+    if (!ok) {
+        LOGE("Could not start recorder thread");
+        goto error_stream_cond_destroy;
+    }
 
     return true;
+
+error_stream_cond_destroy:
+    sc_cond_destroy(&recorder->stream_cond);
+error_queue_cond_destroy:
+    sc_cond_destroy(&recorder->queue_cond);
+error_mutex_destroy:
+    sc_mutex_destroy(&recorder->mutex);
+error_free_filename:
+    free(recorder->filename);
+
+    return false;
+}
+
+void
+sc_recorder_stop(struct sc_recorder *recorder) {
+    sc_mutex_lock(&recorder->mutex);
+    recorder->stopped = true;
+    sc_cond_signal(&recorder->queue_cond);
+    sc_cond_signal(&recorder->stream_cond);
+    sc_mutex_unlock(&recorder->mutex);
+}
+
+void
+sc_recorder_join(struct sc_recorder *recorder) {
+    sc_thread_join(&recorder->thread, NULL);
 }
 
 void
 sc_recorder_destroy(struct sc_recorder *recorder) {
+    sc_cond_destroy(&recorder->stream_cond);
+    sc_cond_destroy(&recorder->queue_cond);
+    sc_mutex_destroy(&recorder->mutex);
     free(recorder->filename);
 }

app/src/recorder.h

@@ -20,34 +20,63 @@ struct sc_record_packet {
 struct sc_recorder_queue SC_QUEUE(struct sc_record_packet);
 
 struct sc_recorder {
-    struct sc_packet_sink packet_sink; // packet sink trait
+    struct sc_packet_sink video_packet_sink;
+    struct sc_packet_sink audio_packet_sink;
+
+    /* The audio flag is unprotected:
+     *  - it is initialized from sc_recorder_init() from the main thread;
+     *  - it may be reset once from the recorder thread if the audio is
+     *    disabled dynamically.
+     *
+     * Therefore, once the recorder thread is started, only the recorder thread
+     * may access it without data races.
+     */
+    bool audio;
 
     char *filename;
     enum sc_record_format format;
     AVFormatContext *ctx;
     struct sc_size declared_frame_size;
-    bool header_written;
-
-    uint64_t pts_origin;
 
     sc_thread thread;
     sc_mutex mutex;
     sc_cond queue_cond;
-    bool stopped; // set on recorder_close()
-    bool failed; // set on packet write failure
-    struct sc_recorder_queue queue;
+    // set on sc_recorder_stop(), packet_sink close or recording failure
+    bool stopped;
+    struct sc_recorder_queue video_queue;
+    struct sc_recorder_queue audio_queue;
 
-    // we can write a packet only once we received the next one so that we can
-    // set its duration (next_pts - current_pts)
-    // "previous" is only accessed from the recorder thread, so it does not
-    // need to be protected by the mutex
-    struct sc_record_packet *previous;
+    // wake up the recorder thread once the video or audio codec is known
+    sc_cond stream_cond;
+    const AVCodec *video_codec;
+    const AVCodec *audio_codec;
+    // Instead of providing an audio_codec, the demuxer may notify that the
+    // stream is disabled if the device could not capture audio
+    bool audio_disabled;
+
+    int video_stream_index;
+    int audio_stream_index;
+
+    const struct sc_recorder_callbacks *cbs;
+    void *cbs_userdata;
+};
+
+struct sc_recorder_callbacks {
+    void (*on_ended)(struct sc_recorder *recorder, bool success,
+                     void *userdata);
 };
 
 bool
 sc_recorder_init(struct sc_recorder *recorder, const char *filename,
-                 enum sc_record_format format,
-                 struct sc_size declared_frame_size);
+                 enum sc_record_format format, bool audio,
+                 struct sc_size declared_frame_size,
+                 const struct sc_recorder_callbacks *cbs, void *cbs_userdata);
+
+void
+sc_recorder_stop(struct sc_recorder *recorder);
+
+void
+sc_recorder_join(struct sc_recorder *recorder);
 
 void
 sc_recorder_destroy(struct sc_recorder *recorder);

app/src/scrcpy.c

@@ -156,9 +156,15 @@ event_loop(struct scrcpy *s) {
     SDL_Event event;
     while (SDL_WaitEvent(&event)) {
         switch (event.type) {
-            case EVENT_STREAM_STOPPED:
+            case SC_EVENT_DEVICE_DISCONNECTED:
                 LOGW("Device disconnected");
                 return SCRCPY_EXIT_DISCONNECTED;
+            case SC_EVENT_DEMUXER_ERROR:
+                LOGE("Demuxer error");
+                return SCRCPY_EXIT_FAILURE;
+            case SC_EVENT_RECORDER_ERROR:
+                LOGE("Recorder error");
+                return SCRCPY_EXIT_FAILURE;
             case SDL_QUIT:
                 LOGD("User requested to quit");
                 return SCRCPY_EXIT_SUCCESS;
@@ -180,10 +186,10 @@ await_for_server(bool *connected) {
                 LOGD("User requested to quit");
                 *connected = false;
                 return true;
-            case EVENT_SERVER_CONNECTION_FAILED:
+            case SC_EVENT_SERVER_CONNECTION_FAILED:
                 LOGE("Server connection failed");
                 return false;
-            case EVENT_SERVER_CONNECTED:
+            case SC_EVENT_SERVER_CONNECTED:
                 LOGD("Server connected");
                 *connected = true;
                 return true;
@@ -234,19 +240,37 @@ av_log_callback(void *avcl, int level, const char *fmt, va_list vl) {
 }
 
 static void
-sc_video_demuxer_on_eos(struct sc_demuxer *demuxer, void *userdata) {
-    (void) demuxer;
+sc_recorder_on_ended(struct sc_recorder *recorder, bool success,
+                     void *userdata) {
+    (void) recorder;
     (void) userdata;
 
-    PUSH_EVENT(EVENT_STREAM_STOPPED);
+    if (!success) {
+        PUSH_EVENT(SC_EVENT_RECORDER_ERROR);
+    }
 }
 
 static void
-sc_audio_demuxer_on_eos(struct sc_demuxer *demuxer, void *userdata) {
+sc_video_demuxer_on_ended(struct sc_demuxer *demuxer, bool eos,
+                          void *userdata) {
     (void) demuxer;
     (void) userdata;
 
-    // TODO
+    if (eos) {
+        PUSH_EVENT(SC_EVENT_DEVICE_DISCONNECTED);
+    } else {
+        PUSH_EVENT(SC_EVENT_DEMUXER_ERROR);
+    }
+}
+
+static void
+sc_audio_demuxer_on_ended(struct sc_demuxer *demuxer, bool eos,
+                          void *userdata) {
+    (void) demuxer;
+    (void) eos;
+    (void) userdata;
+
+    // Contrary to the video demuxer, keep mirroring if only the audio fails
 }
 
 static void
@@ -254,7 +278,7 @@ sc_server_on_connection_failed(struct sc_server *server, void *userdata) {
     (void) server;
     (void) userdata;
 
-    PUSH_EVENT(EVENT_SERVER_CONNECTION_FAILED);
+    PUSH_EVENT(SC_EVENT_SERVER_CONNECTION_FAILED);
 }
 
 static void
@@ -262,7 +286,7 @@ sc_server_on_connected(struct sc_server *server, void *userdata) {
     (void) server;
     (void) userdata;
 
-    PUSH_EVENT(EVENT_SERVER_CONNECTED);
+    PUSH_EVENT(SC_EVENT_SERVER_CONNECTED);
 }
 
 static void
@@ -275,8 +299,9 @@ sc_server_on_disconnected(struct sc_server *server, void *userdata) {
     // event
 }
 
+// Generate a scrcpy id to differentiate multiple running scrcpy instances
 static uint32_t
-scrcpy_generate_uid() {
+scrcpy_generate_scid() {
     struct sc_rand rand;
     sc_rand_init(&rand);
     // Only use 31 bits to avoid issues with signed values on the Java-side
@@ -317,21 +342,23 @@ scrcpy(struct scrcpy_options *options) {
 
     struct sc_acksync *acksync = NULL;
 
-    uint32_t uid = scrcpy_generate_uid();
+    uint32_t scid = scrcpy_generate_scid();
 
     struct sc_server_params params = {
-        .uid = uid,
+        .scid = scid,
         .req_serial = options->serial,
         .select_usb = options->select_usb,
         .select_tcpip = options->select_tcpip,
         .log_level = options->log_level,
         .codec = options->codec,
+        .audio_codec = options->audio_codec,
         .crop = options->crop,
         .port_range = options->port_range,
         .tunnel_host = options->tunnel_host,
         .tunnel_port = options->tunnel_port,
         .max_size = options->max_size,
         .bit_rate = options->bit_rate,
+        .audio_bit_rate = options->audio_bit_rate,
         .max_fps = options->max_fps,
         .lock_video_orientation = options->lock_video_orientation,
         .control = options->control,
@@ -419,10 +446,12 @@ scrcpy(struct scrcpy_options *options) {
 
     struct sc_recorder *rec = NULL;
     if (options->record_filename) {
-        if (!sc_recorder_init(&s->recorder,
-                              options->record_filename,
-                              options->record_format,
-                              info->frame_size)) {
+        static const struct sc_recorder_callbacks recorder_cbs = {
+            .on_ended = sc_recorder_on_ended,
+        };
+        if (!sc_recorder_init(&s->recorder, options->record_filename,
+                              options->record_format, options->audio,
+                              info->frame_size, &recorder_cbs, NULL)) {
             goto end;
         }
         rec = &s->recorder;
@@ -432,16 +461,16 @@ scrcpy(struct scrcpy_options *options) {
     av_log_set_callback(av_log_callback);
 
     static const struct sc_demuxer_callbacks video_demuxer_cbs = {
-        .on_eos = sc_video_demuxer_on_eos,
+        .on_ended = sc_video_demuxer_on_ended,
     };
-    sc_demuxer_init(&s->video_demuxer, s->server.video_socket,
+    sc_demuxer_init(&s->video_demuxer, "video", s->server.video_socket,
                     &video_demuxer_cbs, NULL);
 
     if (options->audio) {
         static const struct sc_demuxer_callbacks audio_demuxer_cbs = {
-            .on_eos = sc_audio_demuxer_on_eos,
+            .on_ended = sc_audio_demuxer_on_ended,
         };
-        sc_demuxer_init(&s->audio_demuxer, s->server.audio_socket,
+        sc_demuxer_init(&s->audio_demuxer, "audio", s->server.audio_socket,
                         &audio_demuxer_cbs, NULL);
     }
 
@@ -450,7 +479,10 @@ scrcpy(struct scrcpy_options *options) {
     }
 
     if (rec) {
-        sc_demuxer_add_sink(&s->video_demuxer, &rec->packet_sink);
+        sc_demuxer_add_sink(&s->video_demuxer, &rec->video_packet_sink);
+        if (options->audio) {
+            sc_demuxer_add_sink(&s->audio_demuxer, &rec->audio_packet_sink);
+        }
     }
 
     struct sc_controller *controller = NULL;
@@ -702,6 +734,9 @@ end:
     if (file_pusher_initialized) {
         sc_file_pusher_stop(&s->file_pusher);
     }
+    if (recorder_initialized) {
+        sc_recorder_stop(&s->recorder);
+    }
     if (screen_initialized) {
         sc_screen_interrupt(&s->screen);
     }
@@ -753,6 +788,7 @@ end:
     }
 
     if (recorder_initialized) {
+        sc_recorder_join(&s->recorder);
         sc_recorder_destroy(&s->recorder);
     }
 

app/src/screen.c

@@ -371,7 +371,7 @@ sc_video_buffer_on_new_frame(struct sc_video_buffer *vb, bool previous_skipped,
     bool need_new_event;
     if (previous_skipped) {
         sc_fps_counter_add_skipped_frame(&screen->fps_counter);
-        // The EVENT_NEW_FRAME triggered for the previous frame will consume
+        // The SC_EVENT_NEW_FRAME triggered for the previous frame will consume
         // this new frame instead, unless the previous event failed
         need_new_event = screen->event_failed;
     } else {
@@ -380,7 +380,7 @@ sc_video_buffer_on_new_frame(struct sc_video_buffer *vb, bool previous_skipped,
 
     if (need_new_event) {
         static SDL_Event new_frame_event = {
-            .type = EVENT_NEW_FRAME,
+            .type = SC_EVENT_NEW_FRAME,
         };
 
         // Post the event on the UI thread
@@ -820,7 +820,7 @@ sc_screen_handle_event(struct sc_screen *screen, SDL_Event *event) {
     bool relative_mode = sc_screen_is_relative_mode(screen);
 
     switch (event->type) {
-        case EVENT_NEW_FRAME: {
+        case SC_EVENT_NEW_FRAME: {
             bool ok = sc_screen_update_frame(screen);
             if (!ok) {
                 LOGW("Frame update failed\n");

app/src/server.c

@@ -165,6 +165,10 @@ sc_server_get_codec_name(enum sc_codec codec) {
             return "h265";
         case SC_CODEC_AV1:
             return "av1";
+        case SC_CODEC_OPUS:
+            return "opus";
+        case SC_CODEC_AAC:
+            return "aac";
         default:
             return NULL;
     }
@@ -213,16 +217,24 @@ execute_server(struct sc_server *server,
         cmd[count++] = p; \
     }
 
-    ADD_PARAM("uid=%08x", params->uid);
+    ADD_PARAM("scid=%08x", params->scid);
     ADD_PARAM("log_level=%s", log_level_to_server_string(params->log_level));
-    ADD_PARAM("bit_rate=%" PRIu32, params->bit_rate);
 
+    if (params->bit_rate) {
+        ADD_PARAM("bit_rate=%" PRIu32, params->bit_rate);
+    }
     if (!params->audio) {
         ADD_PARAM("audio=false");
+    } else if (params->audio_bit_rate) {
+        ADD_PARAM("audio_bit_rate=%" PRIu32, params->audio_bit_rate);
     }
     if (params->codec != SC_CODEC_H264) {
         ADD_PARAM("codec=%s", sc_server_get_codec_name(params->codec));
     }
+    if (params->audio_codec != SC_CODEC_OPUS) {
+        ADD_PARAM("audio_codec=%s",
+            sc_server_get_codec_name(params->audio_codec));
+    }
     if (params->max_size) {
         ADD_PARAM("max_size=%" PRIu16, params->max_size);
     }
@@ -821,7 +833,7 @@ run_server(void *data) {
     LOGD("Device serial: %s", serial);
 
     int r = asprintf(&server->device_socket_name, SC_SOCKET_NAME_PREFIX "%08x",
-                     params->uid);
+                     params->scid);
     if (r == -1) {
         LOG_OOM();
         goto error_connection_failed;

app/src/server.h

@@ -22,10 +22,11 @@ struct sc_server_info {
 };
 
 struct sc_server_params {
-    uint32_t uid;
+    uint32_t scid;
     const char *req_serial;
     enum sc_log_level log_level;
     enum sc_codec codec;
+    enum sc_codec audio_codec;
     const char *crop;
     const char *codec_options;
     const char *encoder_name;
@@ -34,6 +35,7 @@ struct sc_server_params {
     uint16_t tunnel_port;
     uint16_t max_size;
     uint32_t bit_rate;
+    uint32_t audio_bit_rate;
     uint16_t max_fps;
     int8_t lock_video_orientation;
     bool control;

app/src/trait/packet_sink.h

@@ -19,9 +19,20 @@ struct sc_packet_sink {
 };
 
 struct sc_packet_sink_ops {
+    /* The codec instance is static, it is valid until the end of the program */
     bool (*open)(struct sc_packet_sink *sink, const AVCodec *codec);
     void (*close)(struct sc_packet_sink *sink);
     bool (*push)(struct sc_packet_sink *sink, const AVPacket *packet);
+
+    /*/
+     * Called when the input stream has been disabled at runtime.
+     *
+     * If it is called, then open(), close() and push() will never be called.
+     *
+     * It is useful to notify the recorder that the requested audio stream has
+     * finally been disabled because the device could not capture it.
+     */
+    void (*disable)(struct sc_packet_sink *sink);
 };
 
 #endif

app/src/usb/scrcpy_otg.c

@@ -22,7 +22,7 @@ sc_usb_on_disconnected(struct sc_usb *usb, void *userdata) {
     (void) userdata;
 
     SDL_Event event;
-    event.type = EVENT_USB_DEVICE_DISCONNECTED;
+    event.type = SC_EVENT_USB_DEVICE_DISCONNECTED;
     int ret = SDL_PushEvent(&event);
     if (ret < 0) {
         LOGE("Could not post USB disconnection event: %s", SDL_GetError());
@@ -34,7 +34,7 @@ event_loop(struct scrcpy_otg *s) {
     SDL_Event event;
     while (SDL_WaitEvent(&event)) {
         switch (event.type) {
-            case EVENT_USB_DEVICE_DISCONNECTED:
+            case SC_EVENT_USB_DEVICE_DISCONNECTED:
                 LOGW("Device disconnected");
                 return SCRCPY_EXIT_DISCONNECTED;
             case SDL_QUIT:

server/src/main/java/com/genymobile/scrcpy/AudioCodec.java

@@ -3,7 +3,8 @@ package com.genymobile.scrcpy;
 import android.media.MediaFormat;
 
 public enum AudioCodec implements Codec {
-    OPUS(0x6f_70_75_73, "opus", MediaFormat.MIMETYPE_AUDIO_OPUS);
+    OPUS(0x6f_70_75_73, "opus", MediaFormat.MIMETYPE_AUDIO_OPUS),
+    AAC(0x00_61_61_63, "aac", MediaFormat.MIMETYPE_AUDIO_AAC);
 
     private final int id; // 4-byte ASCII representation of the name
     private final String name;
@@ -17,7 +18,7 @@ public enum AudioCodec implements Codec {
 
     @Override
     public Type getType() {
-        return Type.VIDEO;
+        return Type.AUDIO;
     }
 
     @Override

server/src/main/java/com/genymobile/scrcpy/AudioEncoder.java

@@ -11,18 +11,16 @@ import android.media.MediaRecorder;
 import android.os.Build;
 import android.os.Handler;
 import android.os.HandlerThread;
-import android.os.SystemClock;
 
 import java.io.IOException;
 import java.nio.ByteBuffer;
 import java.util.concurrent.ArrayBlockingQueue;
 import java.util.concurrent.BlockingQueue;
-import java.util.concurrent.atomic.AtomicBoolean;
 
 public final class AudioEncoder {
 
     private static class InputTask {
-        final int index;
+        private final int index;
 
         InputTask(int index) {
             this.index = index;
@@ -30,8 +28,8 @@ public final class AudioEncoder {
     }
 
     private static class OutputTask {
-        final int index;
-        final MediaCodec.BufferInfo bufferInfo;
+        private final int index;
+        private final MediaCodec.BufferInfo bufferInfo;
 
         OutputTask(int index, MediaCodec.BufferInfo bufferInfo) {
             this.index = index;
@@ -39,19 +37,18 @@ public final class AudioEncoder {
         }
     }
 
-    private static final String MIMETYPE = MediaFormat.MIMETYPE_AUDIO_OPUS;
     private static final int SAMPLE_RATE = 48000;
     private static final int CHANNELS = 2;
-    private static final int BIT_RATE = 128000;
 
-    private static int BUFFER_MS = 15; // milliseconds
+    private static final int BUFFER_MS = 15; // milliseconds
     private static final int BUFFER_SIZE = SAMPLE_RATE * CHANNELS * BUFFER_MS / 1000;
 
+    private final Streamer streamer;
+    private final int bitRate;
+
     private AudioRecord recorder;
     private MediaCodec mediaCodec;
 
-    private final AtomicBoolean cleanUpDone = new AtomicBoolean(false);
-
     // Capacity of 64 is in practice "infinite" (it is limited by the number of available MediaCodec buffers, typically 4).
     // So many pending tasks would lead to an unacceptable delay anyway.
     private final BlockingQueue<InputTask> inputTasks = new ArrayBlockingQueue<>(64);
@@ -65,6 +62,11 @@ public final class AudioEncoder {
 
     private boolean ended;
 
+    public AudioEncoder(Streamer streamer, int bitRate) {
+        this.streamer = streamer;
+        this.bitRate = bitRate;
+    }
+
     private static AudioFormat createAudioFormat() {
         AudioFormat.Builder builder = new AudioFormat.Builder();
         builder.setEncoding(AudioFormat.ENCODING_PCM_16BIT);
@@ -87,10 +89,10 @@ public final class AudioEncoder {
         return builder.build();
     }
 
-    private static MediaFormat createFormat() {
+    private static MediaFormat createFormat(String mimeType, int bitRate) {
         MediaFormat format = new MediaFormat();
-        format.setString(MediaFormat.KEY_MIME, MIMETYPE);
-        format.setInteger(MediaFormat.KEY_BIT_RATE, BIT_RATE);
+        format.setString(MediaFormat.KEY_MIME, mimeType);
+        format.setInteger(MediaFormat.KEY_BIT_RATE, bitRate);
         format.setInteger(MediaFormat.KEY_CHANNEL_COUNT, CHANNELS);
         format.setInteger(MediaFormat.KEY_SAMPLE_RATE, SAMPLE_RATE);
         return format;
@@ -142,11 +144,13 @@ public final class AudioEncoder {
     }
 
     private void outputThread() throws IOException, InterruptedException {
+        streamer.writeHeader();
+
         while (!Thread.currentThread().isInterrupted()) {
             OutputTask task = outputTasks.take();
             ByteBuffer buffer = mediaCodec.getOutputBuffer(task.index);
             try {
-                Ln.i("Audio packet [pts=" + task.bufferInfo.presentationTimeUs + "] " + buffer.remaining() + " bytes");
+                streamer.writePacket(buffer, task.bufferInfo);
             } finally {
                 mediaCodec.releaseOutputBuffer(task.index, false);
             }
@@ -158,7 +162,7 @@ public final class AudioEncoder {
             try {
                 encode();
             } catch (IOException e) {
-                // this is expected on close
+                Ln.e("Audio encoding error", e);
             } finally {
                 Ln.d("Audio encoder stopped");
             }
@@ -196,25 +200,32 @@ public final class AudioEncoder {
 
     @TargetApi(Build.VERSION_CODES.M)
     public void encode() throws IOException {
-        mediaCodec = MediaCodec.createEncoderByType(MIMETYPE); // may throw IOException
-
         try {
-            recorder = createAudioRecord();
+            try {
+                String mimeType = streamer.getCodec().getMimeType();
 
-            mediaCodecThread = new HandlerThread("AudioEncoder");
-            mediaCodecThread.start();
+                mediaCodec = MediaCodec.createEncoderByType(mimeType); // may throw IOException
+                recorder = createAudioRecord();
 
-            MediaFormat format = createFormat();
-            mediaCodec.setCallback(new EncoderCallback(), new Handler(mediaCodecThread.getLooper()));
-            mediaCodec.configure(format, null, null, MediaCodec.CONFIGURE_FLAG_ENCODE);
+                mediaCodecThread = new HandlerThread("AudioEncoder");
+                mediaCodecThread.start();
 
-            recorder.startRecording();
+                MediaFormat format = createFormat(mimeType, bitRate);
+                mediaCodec.setCallback(new EncoderCallback(), new Handler(mediaCodecThread.getLooper()));
+                mediaCodec.configure(format, null, null, MediaCodec.CONFIGURE_FLAG_ENCODE);
+
+                recorder.startRecording();
+            } catch (Throwable e) {
+                // Notify the client that the audio could not be captured
+                streamer.writeDisableStream();
+                throw e;
+            }
 
             inputThread = new Thread(() -> {
                 try {
                     inputThread();
                 } catch (IOException | InterruptedException e) {
-                    // this is expected on close
+                    Ln.e("Audio capture error", e);
                 } finally {
                     end();
                 }
@@ -223,8 +234,13 @@ public final class AudioEncoder {
             outputThread = new Thread(() -> {
                 try {
                     outputThread();
-                } catch (IOException | InterruptedException e) {
+                } catch (InterruptedException e) {
                     // this is expected on close
+                } catch (IOException e) {
+                    // Broken pipe is expected on close, because the socket is closed by the client
+                    if (!IO.isBrokenPipe(e)) {
+                        Ln.e("Audio encoding error", e);
+                    }
                 } finally {
                     end();
                 }
@@ -234,7 +250,9 @@ public final class AudioEncoder {
             inputThread.start();
             outputThread.start();
         } catch (Throwable e) {
-            mediaCodec.release();
+            if (mediaCodec != null) {
+                mediaCodec.release();
+            }
             if (recorder != null) {
                 recorder.release();
             }

server/src/main/java/com/genymobile/scrcpy/Codec.java

@@ -2,10 +2,16 @@ package com.genymobile.scrcpy;
 
 public interface Codec {
 
-    enum Type {VIDEO}
+    enum Type {
+        VIDEO,
+        AUDIO,
+    }
 
     Type getType();
+
     int getId();
+
     String getName();
+
     String getMimeType();
 }

server/src/main/java/com/genymobile/scrcpy/DesktopConnection.java

@@ -51,17 +51,17 @@ public final class DesktopConnection implements Closeable {
         return localSocket;
     }
 
-    private static String getSocketName(int uid) {
-        if (uid == -1) {
-            // If no UID is set, use "scrcpy" to simplify using scrcpy-server alone
+    private static String getSocketName(int scid) {
+        if (scid == -1) {
+            // If no SCID is set, use "scrcpy" to simplify using scrcpy-server alone
             return SOCKET_NAME_PREFIX;
         }
 
-        return SOCKET_NAME_PREFIX + String.format("_%08x", uid);
+        return SOCKET_NAME_PREFIX + String.format("_%08x", scid);
     }
 
-    public static DesktopConnection open(int uid, boolean tunnelForward, boolean audio, boolean control, boolean sendDummyByte) throws IOException {
-        String socketName = getSocketName(uid);
+    public static DesktopConnection open(int scid, boolean tunnelForward, boolean audio, boolean control, boolean sendDummyByte) throws IOException {
+        String socketName = getSocketName(scid);
 
         LocalSocket videoSocket = null;
         LocalSocket audioSocket = null;

server/src/main/java/com/genymobile/scrcpy/IO.java

@@ -48,4 +48,9 @@ public final class IO {
         }
         return builder.toString();
     }
+
+    public static boolean isBrokenPipe(IOException e) {
+        Throwable cause = e.getCause();
+        return cause instanceof ErrnoException && ((ErrnoException) cause).errno == OsConstants.EPIPE;
+    }
 }

server/src/main/java/com/genymobile/scrcpy/Options.java

@@ -7,11 +7,13 @@ import java.util.List;
 public class Options {
 
     private Ln.Level logLevel = Ln.Level.DEBUG;
-    private int uid = -1; // 31-bit non-negative value, or -1
+    private int scid = -1; // 31-bit non-negative value, or -1
     private boolean audio = true;
     private int maxSize;
     private VideoCodec codec = VideoCodec.H264;
+    private AudioCodec audioCodec = AudioCodec.OPUS;
     private int bitRate = 8000000;
+    private int audioBitRate = 196000;
     private int maxFps;
     private int lockVideoOrientation = -1;
     private boolean tunnelForward;
@@ -42,12 +44,12 @@ public class Options {
         this.logLevel = logLevel;
     }
 
-    public int getUid() {
-        return uid;
+    public int getScid() {
+        return scid;
     }
 
-    public void setUid(int uid) {
-        this.uid = uid;
+    public void setScid(int scid) {
+        this.scid = scid;
     }
 
     public boolean getAudio() {
@@ -74,6 +76,14 @@ public class Options {
         this.codec = codec;
     }
 
+    public AudioCodec getAudioCodec() {
+        return audioCodec;
+    }
+
+    public void setAudioCodec(AudioCodec audioCodec) {
+        this.audioCodec = audioCodec;
+    }
+
     public int getBitRate() {
         return bitRate;
     }
@@ -82,6 +92,14 @@ public class Options {
         this.bitRate = bitRate;
     }
 
+    public int getAudioBitRate() {
+        return audioBitRate;
+    }
+
+    public void setAudioBitRate(int audioBitRate) {
+        this.audioBitRate = audioBitRate;
+    }
+
     public int getMaxFps() {
         return maxFps;
     }

server/src/main/java/com/genymobile/scrcpy/ScreenEncoder.java

@@ -42,8 +42,8 @@ public class ScreenEncoder implements Device.RotationListener {
     private boolean firstFrameSent;
     private int consecutiveErrors;
 
-    public ScreenEncoder(Device device, Streamer streamer, int bitRate, int maxFps, List<CodecOption> codecOptions,
-            String encoderName, boolean downsizeOnError) {
+    public ScreenEncoder(Device device, Streamer streamer, int bitRate, int maxFps, List<CodecOption> codecOptions, String encoderName,
+            boolean downsizeOnError) {
         this.device = device;
         this.streamer = streamer;
         this.bitRate = bitRate;

server/src/main/java/com/genymobile/scrcpy/Server.java

@@ -66,7 +66,7 @@ public final class Server {
 
         Thread initThread = startInitThread(options);
 
-        int uid = options.getUid();
+        int scid = options.getScid();
         boolean tunnelForward = options.isTunnelForward();
         boolean control = options.getControl();
         boolean audio = options.getAudio();
@@ -93,7 +93,7 @@ public final class Server {
             Workarounds.fillAppInfo();
         }
 
-        try (DesktopConnection connection = DesktopConnection.open(uid, tunnelForward, audio, control, sendDummyByte)) {
+        try (DesktopConnection connection = DesktopConnection.open(scid, tunnelForward, audio, control, sendDummyByte)) {
             VideoCodec codec = options.getCodec();
             if (options.getSendDeviceMeta()) {
                 Size videoSize = device.getScreenInfo().getVideoSize();
@@ -111,7 +111,9 @@ public final class Server {
 
             AudioEncoder audioEncoder = null;
             if (audio) {
-                audioEncoder = new AudioEncoder();
+                Streamer audioStreamer = new Streamer(connection.getAudioFd(), options.getAudioCodec(), options.getSendCodecId(),
+                        options.getSendFrameMeta());
+                audioEncoder = new AudioEncoder(audioStreamer, options.getAudioBitRate());
                 audioEncoder.start();
             }
 
@@ -122,7 +124,10 @@ public final class Server {
                 // synchronous
                 screenEncoder.streamScreen();
             } catch (IOException e) {
-                // this is expected on close
+                // Broken pipe is expected on close, because the socket is closed by the client
+                if (!IO.isBrokenPipe(e)) {
+                    Ln.e("Video encoding error", e);
+                }
             } finally {
                 Ln.d("Screen streaming stopped");
                 initThread.interrupt();
@@ -154,6 +159,7 @@ public final class Server {
         return thread;
     }
 
+    @SuppressWarnings("MethodLength")
     private static Options createOptions(String... args) {
         if (args.length < 1) {
             throw new IllegalArgumentException("Missing client version");
@@ -176,12 +182,12 @@ public final class Server {
             String key = arg.substring(0, equalIndex);
             String value = arg.substring(equalIndex + 1);
             switch (key) {
-                case "uid":
-                    int uid = Integer.parseInt(value, 0x10);
-                    if (uid < -1) {
-                        throw new IllegalArgumentException("uid may not be negative (except -1 for 'none'): " + uid);
+                case "scid":
+                    int scid = Integer.parseInt(value, 0x10);
+                    if (scid < -1) {
+                        throw new IllegalArgumentException("scid may not be negative (except -1 for 'none'): " + scid);
                     }
-                    options.setUid(uid);
+                    options.setScid(scid);
                     break;
                 case "log_level":
                     Ln.Level level = Ln.Level.valueOf(value.toUpperCase(Locale.ENGLISH));
@@ -198,6 +204,13 @@ public final class Server {
                     }
                     options.setCodec(codec);
                     break;
+                case "audio_codec":
+                    AudioCodec audioCodec = AudioCodec.findByName(value);
+                    if (audioCodec == null) {
+                        throw new IllegalArgumentException("Audio codec " + value + " not supported");
+                    }
+                    options.setAudioCodec(audioCodec);
+                    break;
                 case "max_size":
                     int maxSize = Integer.parseInt(value) & ~7; // multiple of 8
                     options.setMaxSize(maxSize);
@@ -206,6 +219,10 @@ public final class Server {
                     int bitRate = Integer.parseInt(value);
                     options.setBitRate(bitRate);
                     break;
+                case "audio_bit_rate":
+                    int audioBitRate = Integer.parseInt(value);
+                    options.setAudioBitRate(audioBitRate);
+                    break;
                 case "max_fps":
                     int maxFps = Integer.parseInt(value);
                     options.setMaxFps(maxFps);

server/src/main/java/com/genymobile/scrcpy/Streamer.java

@@ -11,6 +11,8 @@ public final class Streamer {
     private static final long PACKET_FLAG_CONFIG = 1L << 63;
     private static final long PACKET_FLAG_KEY_FRAME = 1L << 62;
 
+    private static final long AOPUSHDR = 0x5244485355504F41L; // "AOPUSHDR" in ASCII (little-endian)
+
     private final FileDescriptor fd;
     private final Codec codec;
     private final boolean sendCodecId;
@@ -38,7 +40,17 @@ public final class Streamer {
         }
     }
 
+    public void writeDisableStream() throws IOException {
+        // Writing 0 (32-bit) as codec-id means that the device disables the stream (because it could not capture)
+        byte[] zeros = new byte[4];
+        IO.writeFully(fd, zeros, 0, zeros.length);
+    }
+
     public void writePacket(ByteBuffer codecBuffer, MediaCodec.BufferInfo bufferInfo) throws IOException {
+        if (codec == AudioCodec.OPUS) {
+            fixOpusConfigPacket(codecBuffer, bufferInfo);
+        }
+
         if (sendFrameMeta) {
             writeFrameMeta(fd, bufferInfo, codecBuffer.remaining());
         }
@@ -64,4 +76,46 @@ public final class Streamer {
         headerBuffer.flip();
         IO.writeFully(fd, headerBuffer);
     }
+
+    private static void fixOpusConfigPacket(ByteBuffer buffer, MediaCodec.BufferInfo bufferInfo) throws IOException {
+        // Here is an example of the config packet received for an OPUS stream:
+        //
+        // 00000000  41 4f 50 55 53 48 44 52  13 00 00 00 00 00 00 00  |AOPUSHDR........|
+        // -------------- BELOW IS THE PART WE MUST PUT AS EXTRADATA  -------------------
+        // 00000010  4f 70 75 73 48 65 61 64  01 01 38 01 80 bb 00 00  |OpusHead..8.....|
+        // 00000020  00 00 00                                          |...             |
+        // ------------------------------------------------------------------------------
+        // 00000020           41 4f 50 55 53  44 4c 59 08 00 00 00 00  |   AOPUSDLY.....|
+        // 00000030  00 00 00 a0 2e 63 00 00  00 00 00 41 4f 50 55 53  |.....c.....AOPUS|
+        // 00000040  50 52 4c 08 00 00 00 00  00 00 00 00 b4 c4 04 00  |PRL.............|
+        // 00000050  00 00 00                                          |...|
+        //
+        // Each "section" is prefixed by a 64-bit ID and a 64-bit length.
+
+        boolean isConfig = (bufferInfo.flags & MediaCodec.BUFFER_FLAG_CODEC_CONFIG) != 0;
+        if (!isConfig) {
+            return;
+        }
+
+        while (buffer.remaining() >= 16) {
+            long id = buffer.getLong();
+            long sizeLong = buffer.getLong();
+            if (sizeLong < 0 || sizeLong >= 0x7FFFFFFF) {
+                throw new IOException("Invalid block size in OPUS header: " + sizeLong);
+            }
+            int size = (int) sizeLong;
+            if (id == AOPUSHDR) {
+                if (buffer.remaining() < size) {
+                    throw new IOException("Not enough data in OPUS header (invalid size: " + size + ")");
+                }
+                // Set the buffer to point to the OPUS header slice
+                buffer.limit(buffer.position() + size);
+                return;
+            }
+
+            buffer.position(buffer.position() + size);
+        }
+
+        throw new IOException("OPUS header not found");
+    }
 }

server/src/main/java/com/genymobile/scrcpy/Workarounds.java

@@ -2,7 +2,6 @@ package com.genymobile.scrcpy;
 
 import android.annotation.SuppressLint;
 import android.app.Application;
-import android.app.Instrumentation;
 import android.content.ContextWrapper;
 import android.content.pm.ApplicationInfo;
 import android.os.Looper;