Share desktop publisher microphone with radio while preserving stream format

tools/chatgui/CMakeLists.txt

@@ -162,8 +162,10 @@ if(FFMPEG_FOUND)
 endif()
 add_library(chatgui_camera STATIC src/call_camera_capture.cpp src/call_camera_worker.cpp src/call_camera_source.h)
 target_link_libraries(chatgui_camera PUBLIC ${QT_LIBS} utun)
+add_library(chatgui_audio_capture STATIC src/audio_capture.cpp)
+target_link_libraries(chatgui_audio_capture PUBLIC ${QT_LIBS} utun)
 add_library(chatgui_stream_publish STATIC src/stream_publish_engine.cpp src/stream_publish_window.cpp src/call_video_engine.cpp)
-target_link_libraries(chatgui_stream_publish PUBLIC chatgui_live_video chatgui_camera)
+target_link_libraries(chatgui_stream_publish PUBLIC chatgui_live_video chatgui_camera chatgui_audio_capture)
 target_include_directories(chatgui_stream_publish PRIVATE ${CMAKE_SOURCE_DIR}/../../lib/libopus/include)
 set(GUI_CAPTURE_ENABLED 0)
 if(QT_PACKAGE STREQUAL "Qt6" AND FFMPEG_FOUND)
@@ -635,7 +637,7 @@ if(CMAKE_SYSTEM_NAME STREQUAL "Linux")
         src/audiorecorder.cpp src/call_audio_engine.cpp src/radio_audio_engine.cpp
         transport/miniaudio_impl.c transport/audio_diagnostics.cpp transport/config_updater.cpp)
     target_include_directories(test_audio_recovery PRIVATE ${UTUN_INCLUDE_DIRS})
-    target_link_libraries(test_audio_recovery PRIVATE ${QT_LIBS} utun_voice utun pthread dl)
+    target_link_libraries(test_audio_recovery PRIVATE chatgui_audio_capture ${QT_LIBS} utun_voice utun pthread dl)
     target_link_options(test_audio_recovery PRIVATE "-Wl,--wrap=ma_context_init" "-Wl,--wrap=ma_context_uninit"
         "-Wl,--wrap=ma_device_init" "-Wl,--wrap=ma_device_start" "-Wl,--wrap=ma_device_uninit" "-Wl,--wrap=audio_context_failed"
         "-Wl,--wrap=call_send_media")
@@ -646,7 +648,7 @@ if(CMAKE_SYSTEM_NAME STREQUAL "Linux")
         src/sound_manager.cpp src/voice_audio_io.cpp src/audio_device.cpp
         transport/miniaudio_impl.c transport/audio_diagnostics.cpp transport/config_updater.cpp ${RADIO_MEMBER_TEST_SOURCES})
     target_include_directories(test_radio_panel PRIVATE ${UTUN_INCLUDE_DIRS})
-    target_link_libraries(test_radio_panel PRIVATE ${QT_LIBS} ${QT_TEST_LIB} utun_voice utun pthread dl ${X11_LIBRARIES})
+    target_link_libraries(test_radio_panel PRIVATE chatgui_audio_capture ${QT_LIBS} ${QT_TEST_LIB} utun_voice utun pthread dl ${X11_LIBRARIES})
     target_link_options(test_radio_panel PRIVATE "-Wl,--wrap=ma_context_init" "-Wl,--wrap=ma_device_start"
         "-Wl,--wrap=radio_talk_begin" "-Wl,--wrap=radio_talk_end" "-Wl,--wrap=radio_talk_send"
         "-Wl,--wrap=silero_vad_create_default" "-Wl,--wrap=silero_vad_destroy" "-Wl,--wrap=silero_vad_reset"

tools/chatgui/src/stream_publish_engine.cpp

@@ -307,36 +307,21 @@ void StreamPublishEngine::stop() {
 }
 
 bool StreamPublishEngine::openMicrophone() {
-    m_capture.close();
-    m->audioAnchor = -1; m->audioFrames = 0;
-    auto config = this->config(); auto* context = SoundManager::instance()->context();
-    if (!context) { DEBUG_ERROR(DEBUG_CATEGORY_STREAM, "desktop stream audio context unavailable"); m_capture.retryLater(); return false; }
-    ma_device_id device; bool selected = SoundManager::deviceId(config.microphone, &device);
-    if (!config.microphone.isEmpty() && !selected) {
-        DEBUG_ERROR(DEBUG_CATEGORY_STREAM, "desktop selected microphone ID is unavailable"); m_capture.retryLater(); return false;
-    }
-    for (int channels : {2, 1}) {
-        ma_device_config capture = ma_device_config_init(ma_device_type_capture);
-        capture.capture.format = ma_format_s16; capture.capture.channels = channels; capture.sampleRate = 48000;
-        capture.capture.pDeviceID = selected ? &device : nullptr;
-        auto callback = [this](const void* in, void*, unsigned frames) { captureAudio(in, frames); };
-        if (!m_capture.init(context, capture, "stream-capture", m->id, callback)) continue;
-        m->channels = channels;
-        if (!m_capture.start()) continue;
-        int native = m_capture.device()->capture.internalChannels;
+    return m_capture.open(config().microphone, AudioCapture::Role::Stream, 2, m->id, [this](int channels, int native) {
+        m->audioAnchor = -1; m->audioFrames = 0; m->channels = channels;
         if (native == 1) DEBUG_WARN(DEBUG_CATEGORY_STREAM, "desktop stream input is mono; stereo transport duplicates the input");
-        DEBUG_INFO(DEBUG_CATEGORY_STREAM, "desktop microphone opened requested=%d backend_channels=%d rate=%u", channels, native, m_capture.device()->sampleRate);
+        DEBUG_INFO(DEBUG_CATEGORY_STREAM, "desktop microphone opened requested=%d backend_channels=%d rate=48000", channels, native);
         emit audioChannelsChanged(native == 1 ? 1 : channels); return true;
-    }
-    DEBUG_ERROR(DEBUG_CATEGORY_STREAM, "desktop stream cannot open microphone: %s", qPrintable(config.microphone)); return false;
+    }, [] {}, [this](const AudioCapture::Frame& frame) { captureAudio(frame); });
 }
 
 /* Hardware callback копирует максимум 20ms за элемент, без codec/DSP/allocations. */
-void StreamPublishEngine::captureAudio(const void* input, unsigned frames) {
-    const auto* pcm = static_cast<const int16_t*>(input);
+void StreamPublishEngine::captureAudio(const AudioCapture::Frame& frame) {
+    const auto* pcm = frame.pcm;
+    unsigned frames = frame.frames;
     if (!m->running) return;
-    int64_t first = m_capture.callbackTimeUs() - m->origin;
-    if (m_capture.timingMeasured() || m->audioAnchor < 0 || m_capture.captureMissing()) {
+    int64_t first = frame.timeUs - m->origin;
+    if (frame.measured || m->audioAnchor < 0 || frame.missing) {
         m->audioAnchor = first; m->audioFrames = 0;
     }
     first = m->audioAnchor + (int64_t)m->audioFrames * 1000000 / 48000; m->audioFrames += frames;
@@ -540,8 +525,9 @@ void StreamPublishEngine::tick() {
                   (unsigned long long)m->sourceGeneration, (unsigned long long)m->generation, (unsigned long long)m->confirmedGeneration,
                   (unsigned long long)m->capturedFrames, (long long)m->framePts);
           } }
-        if (m->wanted.audio && m_capture.needsRecovery()) { m_capture.close(); m_capture.retryLater(); }
-        if (m->wanted.audio && m_capture.retryReady()) openMicrophone();
+        if (m->wanted.audio) {
+            if (!m_capture.subscribed()) openMicrophone(); else m_capture.recover();
+        }
     }
     unsigned dropped = m->audioDropped.exchange(0);
     if (dropped) DEBUG_WARN(DEBUG_CATEGORY_STREAM, "desktop audio capture discarded blocks=%u", dropped);

tools/chatgui/src/stream_publish_engine.h

@@ -9,7 +9,7 @@
 #include <QList>
 #include <memory>
 #include "live_video_encoder.h"
-#include "audio_device.h"
+#include "audio_capture.h"
 #include "audio_capture_lease.h"
 #include "../../../src/stream/stream_publisher.h"
 
@@ -53,7 +53,7 @@ signals:
 private:
     struct Impl;
     std::unique_ptr<Impl> m;
-    AudioDevice m_capture;
+    AudioCapture m_capture;
     AudioCaptureLease m_microphone{AudioCaptureLease::Mode::Exclusive};
     QTimer m_timer;
     QTimer m_previewTimer;
@@ -65,5 +65,5 @@ private:
     void tick();
     void videoLoop();
     void audioLoop();
-    void captureAudio(const void* pcm, unsigned frames);
+    void captureAudio(const AudioCapture::Frame& frame);
 };