Merge pull request #465 from trilene/call-devices

Add Duplex call devices

12 files changed, 530 insertions, 407 deletions

diff --git a/CMakeLists.txt b/CMakeLists.txt
index c9e29998..34330147 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -298,6 +298,7 @@ set(SRC_FILES
 	src/AvatarProvider.cpp
 	src/BlurhashProvider.cpp
 	src/Cache.cpp
+	src/CallDevices.cpp
 	src/CallManager.cpp
 	src/ChatPage.cpp
 	src/ColorImageProvider.cpp
@@ -512,6 +513,7 @@ qt5_wrap_cpp(MOC_HEADERS
 	src/AvatarProvider.h
 	src/BlurhashProvider.h
 	src/Cache_p.h
+	src/CallDevices.h
 	src/CallManager.h
 	src/ChatPage.h
 	src/CommunitiesList.h
diff --git a/resources/qml/MessageView.qml b/resources/qml/MessageView.qml
index 29115b00..dafca0f6 100644
--- a/resources/qml/MessageView.qml
+++ b/resources/qml/MessageView.qml
@@ -102,6 +102,7 @@ ListView {
 
                 Avatar {
                     id: messageUserAvatar
+
                     width: avatarSize
                     height: avatarSize
                     url: modelData ? chat.model.avatarUrl(modelData.userId).replace("mxc://", "image://MxcImage/") : ""
@@ -112,8 +113,8 @@ ListView {
 
                 Connections {
                     target: chat.model
-                    onRoomAvatarUrlChanged: { 
-                      messageUserAvatar.url = modelData ? chat.model.avatarUrl(modelData.userId).replace("mxc://", "image://MxcImage/") : ""
+                    onRoomAvatarUrlChanged: {
+                        messageUserAvatar.url = modelData ? chat.model.avatarUrl(modelData.userId).replace("mxc://", "image://MxcImage/") : "";
                     }
                 }
 
diff --git a/resources/qml/UserProfile.qml b/resources/qml/UserProfile.qml
index 37ae6de8..4797a38e 100644
--- a/resources/qml/UserProfile.qml
+++ b/resources/qml/UserProfile.qml
@@ -49,6 +49,7 @@ ApplicationWindow {
 
         Text {
             id: errorText
+
             text: "Error Text"
             color: "red"
             visible: opacity > 0
@@ -58,24 +59,28 @@ ApplicationWindow {
 
         SequentialAnimation {
             id: hideErrorAnimation
+
             running: false
+
             PauseAnimation {
                 duration: 4000
             }
+
             NumberAnimation {
                 target: errorText
                 property: 'opacity'
                 to: 0
                 duration: 1000
             }
+
         }
 
-        Connections{
+        Connections {
             target: profile
             onDisplayError: {
-                errorText.text = errorMessage
-                errorText.opacity = 1
-                hideErrorAnimation.restart()
+                errorText.text = errorMessage;
+                errorText.opacity = 1;
+                hideErrorAnimation.restart();
             }
         }
 
diff --git a/src/CallDevices.cpp b/src/CallDevices.cpp
new file mode 100644
index 00000000..0b9809e5
--- /dev/null
+++ b/src/CallDevices.cpp
@@ -0,0 +1,437 @@
+#include <cstring>
+#include <optional>
+#include <string_view>
+
+#include "CallDevices.h"
+#include "ChatPage.h"
+#include "Logging.h"
+#include "UserSettingsPage.h"
+
+#ifdef GSTREAMER_AVAILABLE
+extern "C"
+{
+#include "gst/gst.h"
+}
+#endif
+
+CallDevices::CallDevices()
+  : QObject()
+{}
+
+#ifdef GSTREAMER_AVAILABLE
+namespace {
+
+struct AudioSource
+{
+        std::string name;
+        GstDevice *device;
+};
+
+struct VideoSource
+{
+        struct Caps
+        {
+                std::string resolution;
+                std::vector<std::string> frameRates;
+        };
+        std::string name;
+        GstDevice *device;
+        std::vector<Caps> caps;
+};
+
+std::vector<AudioSource> audioSources_;
+std::vector<VideoSource> videoSources_;
+
+using FrameRate = std::pair<int, int>;
+std::optional<FrameRate>
+getFrameRate(const GValue *value)
+{
+        if (GST_VALUE_HOLDS_FRACTION(value)) {
+                gint num = gst_value_get_fraction_numerator(value);
+                gint den = gst_value_get_fraction_denominator(value);
+                return FrameRate{num, den};
+        }
+        return std::nullopt;
+}
+
+void
+addFrameRate(std::vector<std::string> &rates, const FrameRate &rate)
+{
+        constexpr double minimumFrameRate = 15.0;
+        if (static_cast<double>(rate.first) / rate.second >= minimumFrameRate)
+                rates.push_back(std::to_string(rate.first) + "/" + std::to_string(rate.second));
+}
+
+void
+setDefaultDevice(bool isVideo)
+{
+        auto settings = ChatPage::instance()->userSettings();
+        if (isVideo && settings->camera().isEmpty()) {
+                const VideoSource &camera = videoSources_.front();
+                settings->setCamera(QString::fromStdString(camera.name));
+                settings->setCameraResolution(
+                  QString::fromStdString(camera.caps.front().resolution));
+                settings->setCameraFrameRate(
+                  QString::fromStdString(camera.caps.front().frameRates.front()));
+        } else if (!isVideo && settings->microphone().isEmpty()) {
+                settings->setMicrophone(QString::fromStdString(audioSources_.front().name));
+        }
+}
+
+void
+addDevice(GstDevice *device)
+{
+        if (!device)
+                return;
+
+        gchar *name  = gst_device_get_display_name(device);
+        gchar *type  = gst_device_get_device_class(device);
+        bool isVideo = !std::strncmp(type, "Video", 5);
+        g_free(type);
+        nhlog::ui()->debug("WebRTC: {} device added: {}", isVideo ? "video" : "audio", name);
+        if (!isVideo) {
+                audioSources_.push_back({name, device});
+                g_free(name);
+                setDefaultDevice(false);
+                return;
+        }
+
+        GstCaps *gstcaps = gst_device_get_caps(device);
+        if (!gstcaps) {
+                nhlog::ui()->debug("WebRTC: unable to get caps for {}", name);
+                g_free(name);
+                return;
+        }
+
+        VideoSource source{name, device, {}};
+        g_free(name);
+        guint nCaps = gst_caps_get_size(gstcaps);
+        for (guint i = 0; i < nCaps; ++i) {
+                GstStructure *structure = gst_caps_get_structure(gstcaps, i);
+                const gchar *name       = gst_structure_get_name(structure);
+                if (!std::strcmp(name, "video/x-raw")) {
+                        gint widthpx, heightpx;
+                        if (gst_structure_get(structure,
+                                              "width",
+                                              G_TYPE_INT,
+                                              &widthpx,
+                                              "height",
+                                              G_TYPE_INT,
+                                              &heightpx,
+                                              nullptr)) {
+                                VideoSource::Caps caps;
+                                caps.resolution =
+                                  std::to_string(widthpx) + "x" + std::to_string(heightpx);
+                                const GValue *value =
+                                  gst_structure_get_value(structure, "framerate");
+                                if (auto fr = getFrameRate(value); fr)
+                                        addFrameRate(caps.frameRates, *fr);
+                                else if (GST_VALUE_HOLDS_FRACTION_RANGE(value)) {
+                                        addFrameRate(
+                                          caps.frameRates,
+                                          *getFrameRate(gst_value_get_fraction_range_min(value)));
+                                        addFrameRate(
+                                          caps.frameRates,
+                                          *getFrameRate(gst_value_get_fraction_range_max(value)));
+                                } else if (GST_VALUE_HOLDS_LIST(value)) {
+                                        guint nRates = gst_value_list_get_size(value);
+                                        for (guint j = 0; j < nRates; ++j) {
+                                                const GValue *rate =
+                                                  gst_value_list_get_value(value, j);
+                                                if (auto fr = getFrameRate(rate); fr)
+                                                        addFrameRate(caps.frameRates, *fr);
+                                        }
+                                }
+                                if (!caps.frameRates.empty())
+                                        source.caps.push_back(std::move(caps));
+                        }
+                }
+        }
+        gst_caps_unref(gstcaps);
+        videoSources_.push_back(std::move(source));
+        setDefaultDevice(true);
+}
+
+#if GST_CHECK_VERSION(1, 18, 0)
+template<typename T>
+bool
+removeDevice(T &sources, GstDevice *device, bool changed)
+{
+        if (auto it = std::find_if(sources.begin(),
+                                   sources.end(),
+                                   [device](const auto &s) { return s.device == device; });
+            it != sources.end()) {
+                nhlog::ui()->debug(std::string("WebRTC: device ") +
+                                     (changed ? "changed: " : "removed: ") + "{}",
+                                   it->name);
+                gst_object_unref(device);
+                sources.erase(it);
+                return true;
+        }
+        return false;
+}
+
+void
+removeDevice(GstDevice *device, bool changed)
+{
+        if (device) {
+                if (removeDevice(audioSources_, device, changed) ||
+                    removeDevice(videoSources_, device, changed))
+                        return;
+        }
+}
+
+gboolean
+newBusMessage(GstBus *bus G_GNUC_UNUSED, GstMessage *msg, gpointer user_data G_GNUC_UNUSED)
+{
+        switch (GST_MESSAGE_TYPE(msg)) {
+        case GST_MESSAGE_DEVICE_ADDED: {
+                GstDevice *device;
+                gst_message_parse_device_added(msg, &device);
+                addDevice(device);
+                emit CallDevices::instance().devicesChanged();
+                break;
+        }
+        case GST_MESSAGE_DEVICE_REMOVED: {
+                GstDevice *device;
+                gst_message_parse_device_removed(msg, &device);
+                removeDevice(device, false);
+                emit CallDevices::instance().devicesChanged();
+                break;
+        }
+        case GST_MESSAGE_DEVICE_CHANGED: {
+                GstDevice *device;
+                GstDevice *oldDevice;
+                gst_message_parse_device_changed(msg, &device, &oldDevice);
+                removeDevice(oldDevice, true);
+                addDevice(device);
+                break;
+        }
+        default:
+                break;
+        }
+        return TRUE;
+}
+#endif
+
+template<typename T>
+std::vector<std::string>
+deviceNames(T &sources, const std::string &defaultDevice)
+{
+        std::vector<std::string> ret;
+        ret.reserve(sources.size());
+        for (const auto &s : sources)
+                ret.push_back(s.name);
+
+        // move default device to top of the list
+        if (auto it = std::find(ret.begin(), ret.end(), defaultDevice); it != ret.end())
+                std::swap(ret.front(), *it);
+
+        return ret;
+}
+
+std::optional<VideoSource>
+getVideoSource(const std::string &cameraName)
+{
+        if (auto it = std::find_if(videoSources_.cbegin(),
+                                   videoSources_.cend(),
+                                   [&cameraName](const auto &s) { return s.name == cameraName; });
+            it != videoSources_.cend()) {
+                return *it;
+        }
+        return std::nullopt;
+}
+
+std::pair<int, int>
+tokenise(std::string_view str, char delim)
+{
+        std::pair<int, int> ret;
+        ret.first  = std::atoi(str.data());
+        auto pos   = str.find_first_of(delim);
+        ret.second = std::atoi(str.data() + pos + 1);
+        return ret;
+}
+
+}
+
+void
+CallDevices::init()
+{
+#if GST_CHECK_VERSION(1, 18, 0)
+        static GstDeviceMonitor *monitor = nullptr;
+        if (!monitor) {
+                monitor       = gst_device_monitor_new();
+                GstCaps *caps = gst_caps_new_empty_simple("audio/x-raw");
+                gst_device_monitor_add_filter(monitor, "Audio/Source", caps);
+                gst_device_monitor_add_filter(monitor, "Audio/Duplex", caps);
+                gst_caps_unref(caps);
+                caps = gst_caps_new_empty_simple("video/x-raw");
+                gst_device_monitor_add_filter(monitor, "Video/Source", caps);
+                gst_device_monitor_add_filter(monitor, "Video/Duplex", caps);
+                gst_caps_unref(caps);
+
+                GstBus *bus = gst_device_monitor_get_bus(monitor);
+                gst_bus_add_watch(bus, newBusMessage, nullptr);
+                gst_object_unref(bus);
+                if (!gst_device_monitor_start(monitor)) {
+                        nhlog::ui()->error("WebRTC: failed to start device monitor");
+                        return;
+                }
+        }
+#endif
+}
+
+void
+CallDevices::refresh()
+{
+#if !GST_CHECK_VERSION(1, 18, 0)
+
+        static GstDeviceMonitor *monitor = nullptr;
+        if (!monitor) {
+                monitor       = gst_device_monitor_new();
+                GstCaps *caps = gst_caps_new_empty_simple("audio/x-raw");
+                gst_device_monitor_add_filter(monitor, "Audio/Source", caps);
+                gst_device_monitor_add_filter(monitor, "Audio/Duplex", caps);
+                gst_caps_unref(caps);
+                caps = gst_caps_new_empty_simple("video/x-raw");
+                gst_device_monitor_add_filter(monitor, "Video/Source", caps);
+                gst_device_monitor_add_filter(monitor, "Video/Duplex", caps);
+                gst_caps_unref(caps);
+        }
+
+        auto clearDevices = [](auto &sources) {
+                std::for_each(
+                  sources.begin(), sources.end(), [](auto &s) { gst_object_unref(s.device); });
+                sources.clear();
+        };
+        clearDevices(audioSources_);
+        clearDevices(videoSources_);
+
+        GList *devices = gst_device_monitor_get_devices(monitor);
+        if (devices) {
+                for (GList *l = devices; l != nullptr; l = l->next)
+                        addDevice(GST_DEVICE_CAST(l->data));
+                g_list_free(devices);
+        }
+        emit devicesChanged();
+#endif
+}
+
+bool
+CallDevices::haveMic() const
+{
+        return !audioSources_.empty();
+}
+
+bool
+CallDevices::haveCamera() const
+{
+        return !videoSources_.empty();
+}
+
+std::vector<std::string>
+CallDevices::names(bool isVideo, const std::string &defaultDevice) const
+{
+        return isVideo ? deviceNames(videoSources_, defaultDevice)
+                       : deviceNames(audioSources_, defaultDevice);
+}
+
+std::vector<std::string>
+CallDevices::resolutions(const std::string &cameraName) const
+{
+        std::vector<std::string> ret;
+        if (auto s = getVideoSource(cameraName); s) {
+                ret.reserve(s->caps.size());
+                for (const auto &c : s->caps)
+                        ret.push_back(c.resolution);
+        }
+        return ret;
+}
+
+std::vector<std::string>
+CallDevices::frameRates(const std::string &cameraName, const std::string &resolution) const
+{
+        if (auto s = getVideoSource(cameraName); s) {
+                if (auto it =
+                      std::find_if(s->caps.cbegin(),
+                                   s->caps.cend(),
+                                   [&](const auto &c) { return c.resolution == resolution; });
+                    it != s->caps.cend())
+                        return it->frameRates;
+        }
+        return {};
+}
+
+GstDevice *
+CallDevices::audioDevice() const
+{
+        std::string name = ChatPage::instance()->userSettings()->microphone().toStdString();
+        if (auto it = std::find_if(audioSources_.cbegin(),
+                                   audioSources_.cend(),
+                                   [&name](const auto &s) { return s.name == name; });
+            it != audioSources_.cend()) {
+                nhlog::ui()->debug("WebRTC: microphone: {}", name);
+                return it->device;
+        } else {
+                nhlog::ui()->error("WebRTC: unknown microphone: {}", name);
+                return nullptr;
+        }
+}
+
+GstDevice *
+CallDevices::videoDevice(std::pair<int, int> &resolution, std::pair<int, int> &frameRate) const
+{
+        auto settings    = ChatPage::instance()->userSettings();
+        std::string name = settings->camera().toStdString();
+        if (auto s = getVideoSource(name); s) {
+                nhlog::ui()->debug("WebRTC: camera: {}", name);
+                resolution = tokenise(settings->cameraResolution().toStdString(), 'x');
+                frameRate  = tokenise(settings->cameraFrameRate().toStdString(), '/');
+                nhlog::ui()->debug(
+                  "WebRTC: camera resolution: {}x{}", resolution.first, resolution.second);
+                nhlog::ui()->debug(
+                  "WebRTC: camera frame rate: {}/{}", frameRate.first, frameRate.second);
+                return s->device;
+        } else {
+                nhlog::ui()->error("WebRTC: unknown camera: {}", name);
+                return nullptr;
+        }
+}
+
+#else
+
+void
+CallDevices::refresh()
+{}
+
+bool
+CallDevices::haveMic() const
+{
+        return false;
+}
+
+bool
+CallDevices::haveCamera() const
+{
+        return false;
+}
+
+std::vector<std::string>
+CallDevices::names(bool, const std::string &) const
+{
+        return {};
+}
+
+std::vector<std::string>
+CallDevices::resolutions(const std::string &) const
+{
+        return {};
+}
+
+std::vector<std::string>
+CallDevices::frameRates(const std::string &, const std::string &) const
+{
+        return {};
+}
+
+#endif
diff --git a/src/CallDevices.h b/src/CallDevices.h
new file mode 100644
index 00000000..2b4129f1
--- /dev/null
+++ b/src/CallDevices.h
@@ -0,0 +1,45 @@
+#pragma once
+
+#include <string>
+#include <utility>
+#include <vector>
+
+#include <QObject>
+
+typedef struct _GstDevice GstDevice;
+
+class CallDevices : public QObject
+{
+        Q_OBJECT
+
+public:
+        static CallDevices &instance()
+        {
+                static CallDevices instance;
+                return instance;
+        }
+
+        void refresh();
+        bool haveMic() const;
+        bool haveCamera() const;
+        std::vector<std::string> names(bool isVideo, const std::string &defaultDevice) const;
+        std::vector<std::string> resolutions(const std::string &cameraName) const;
+        std::vector<std::string> frameRates(const std::string &cameraName,
+                                            const std::string &resolution) const;
+
+signals:
+        void devicesChanged();
+
+private:
+        CallDevices();
+
+        friend class WebRTCSession;
+        void init();
+        GstDevice *audioDevice() const;
+        GstDevice *videoDevice(std::pair<int, int> &resolution,
+                               std::pair<int, int> &frameRate) const;
+
+public:
+        CallDevices(CallDevices const &) = delete;
+        void operator=(CallDevices const &) = delete;
+};
diff --git a/src/CallManager.cpp b/src/CallManager.cpp
index 0841a079..7acd9592 100644
--- a/src/CallManager.cpp
+++ b/src/CallManager.cpp
@@ -7,6 +7,7 @@
 #include <QUrl>
 
 #include "Cache.h"
+#include "CallDevices.h"
 #include "CallManager.h"
 #include "ChatPage.h"
 #include "Logging.h"
@@ -114,21 +115,10 @@ CallManager::CallManager(QObject *parent)
                 emit newCallState();
         });
 
-        connect(&session_, &WebRTCSession::devicesChanged, this, [this]() {
-                if (ChatPage::instance()->userSettings()->microphone().isEmpty()) {
-                        auto mics = session_.getDeviceNames(false, std::string());
-                        if (!mics.empty())
-                                ChatPage::instance()->userSettings()->setMicrophone(
-                                  QString::fromStdString(mics.front()));
-                }
-                if (ChatPage::instance()->userSettings()->camera().isEmpty()) {
-                        auto cameras = session_.getDeviceNames(true, std::string());
-                        if (!cameras.empty())
-                                ChatPage::instance()->userSettings()->setCamera(
-                                  QString::fromStdString(cameras.front()));
-                }
-                emit devicesChanged();
-        });
+        connect(&CallDevices::instance(),
+                &CallDevices::devicesChanged,
+                this,
+                &CallManager::devicesChanged);
 
         connect(&player_,
                 &QMediaPlayer::mediaStatusChanged,
@@ -292,7 +282,7 @@ CallManager::handleEvent(const RoomEvent<CallInvite> &callInviteEvent)
         haveCallInvite_ = true;
         isVideo_        = isVideo;
         inviteSDP_      = callInviteEvent.content.sdp;
-        session_.refreshDevices();
+        CallDevices::instance().refresh();
         emit newInviteState();
 }
 
@@ -409,7 +399,7 @@ CallManager::devices(bool isVideo) const
         const QString &defaultDevice = isVideo ? ChatPage::instance()->userSettings()->camera()
                                                : ChatPage::instance()->userSettings()->microphone();
         std::vector<std::string> devices =
-          session_.getDeviceNames(isVideo, defaultDevice.toStdString());
+          CallDevices::instance().names(isVideo, defaultDevice.toStdString());
         ret.reserve(devices.size());
         std::transform(devices.cbegin(),
                        devices.cend(),
diff --git a/src/CallManager.h b/src/CallManager.h
index 7d388efd..97cffbc8 100644
--- a/src/CallManager.h
+++ b/src/CallManager.h
@@ -8,6 +8,7 @@
 #include <QString>
 #include <QTimer>
 
+#include "CallDevices.h"
 #include "WebRTCSession.h"
 #include "mtx/events/collections.hpp"
 #include "mtx/events/voip.hpp"
@@ -53,7 +54,7 @@ public:
 public slots:
         void sendInvite(const QString &roomid, bool isVideo);
         void syncEvent(const mtx::events::collections::TimelineEvents &event);
-        void refreshDevices() { session_.refreshDevices(); }
+        void refreshDevices() { CallDevices::instance().refresh(); }
         void toggleMicMute();
         void toggleCameraView() { session_.toggleCameraView(); }
         void acceptInvite();
diff --git a/src/ChatPage.cpp b/src/ChatPage.cpp
index d8907740..6d67e6f2 100644
--- a/src/ChatPage.cpp
+++ b/src/ChatPage.cpp
@@ -449,10 +449,9 @@ ChatPage::deleteConfigs()
 {
         QSettings settings;
 
-        if (UserSettings::instance()->profile() != "")
-        {
-            settings.beginGroup("profile");
-            settings.beginGroup(UserSettings::instance()->profile());
+        if (UserSettings::instance()->profile() != "") {
+                settings.beginGroup("profile");
+                settings.beginGroup(UserSettings::instance()->profile());
         }
         settings.beginGroup("auth");
         settings.remove("");
diff --git a/src/UserSettingsPage.cpp b/src/UserSettingsPage.cpp
index d31c8ef9..d4e56b4d 100644
--- a/src/UserSettingsPage.cpp
+++ b/src/UserSettingsPage.cpp
@@ -39,12 +39,12 @@
 #include <QtQml>
 
 #include "Cache.h"
+#include "CallDevices.h"
 #include "Config.h"
 #include "MatrixClient.h"
 #include "Olm.h"
 #include "UserSettingsPage.h"
 #include "Utils.h"
-#include "WebRTCSession.h"
 #include "ui/FlatButton.h"
 #include "ui/ToggleButton.h"
 
@@ -1060,7 +1060,7 @@ UserSettingsPage::UserSettingsPage(QSharedPointer<UserSettings> settings, QWidge
                 [this](const QString &camera) {
                         settings_->setCamera(camera);
                         std::vector<std::string> resolutions =
-                          WebRTCSession::instance().getResolutions(camera.toStdString());
+                          CallDevices::instance().resolutions(camera.toStdString());
                         cameraResolutionCombo_->clear();
                         for (const auto &resolution : resolutions)
                                 cameraResolutionCombo_->addItem(QString::fromStdString(resolution));
@@ -1070,9 +1070,8 @@ UserSettingsPage::UserSettingsPage(QSharedPointer<UserSettings> settings, QWidge
                 static_cast<void (QComboBox::*)(const QString &)>(&QComboBox::currentTextChanged),
                 [this](const QString &resolution) {
                         settings_->setCameraResolution(resolution);
-                        std::vector<std::string> frameRates =
-                          WebRTCSession::instance().getFrameRates(settings_->camera().toStdString(),
-                                                                  resolution.toStdString());
+                        std::vector<std::string> frameRates = CallDevices::instance().frameRates(
+                          settings_->camera().toStdString(), resolution.toStdString());
                         cameraFrameRateCombo_->clear();
                         for (const auto &frameRate : frameRates)
                                 cameraFrameRateCombo_->addItem(QString::fromStdString(frameRate));
@@ -1231,9 +1230,8 @@ UserSettingsPage::showEvent(QShowEvent *)
         timelineMaxWidthSpin_->setValue(settings_->timelineMaxWidth());
         privacyScreenTimeout_->setValue(settings_->privacyScreenTimeout());
 
-        WebRTCSession::instance().refreshDevices();
-        auto mics =
-          WebRTCSession::instance().getDeviceNames(false, settings_->microphone().toStdString());
+        CallDevices::instance().refresh();
+        auto mics = CallDevices::instance().names(false, settings_->microphone().toStdString());
         microphoneCombo_->clear();
         for (const auto &m : mics)
                 microphoneCombo_->addItem(QString::fromStdString(m));
@@ -1241,8 +1239,7 @@ UserSettingsPage::showEvent(QShowEvent *)
         auto cameraResolution = settings_->cameraResolution();
         auto cameraFrameRate  = settings_->cameraFrameRate();
 
-        auto cameras =
-          WebRTCSession::instance().getDeviceNames(true, settings_->camera().toStdString());
+        auto cameras = CallDevices::instance().names(true, settings_->camera().toStdString());
         cameraCombo_->clear();
         for (const auto &c : cameras)
                 cameraCombo_->addItem(QString::fromStdString(c));
diff --git a/src/Utils.cpp b/src/Utils.cpp
index 1b2808b3..f90e5049 100644
--- a/src/Utils.cpp
+++ b/src/Utils.cpp
@@ -70,7 +70,8 @@ utils::replaceEmoji(const QString &body)
         for (auto &code : utf32_string) {
                 if (utils::codepointIsEmoji(code)) {
                         if (!insideFontBlock) {
-                                fmtBody += QString("<font face=\"" + UserSettings::instance()->font() + "\">");
+                                fmtBody += QString("<font face=\"" +
+                                                   UserSettings::instance()->font() + "\">");
                                 insideFontBlock = true;
                         }
 
diff --git a/src/WebRTCSession.cpp b/src/WebRTCSession.cpp
index d306007d..b6d98058 100644
--- a/src/WebRTCSession.cpp
+++ b/src/WebRTCSession.cpp
@@ -35,6 +35,7 @@ using webrtc::State;
 
 WebRTCSession::WebRTCSession()
   : QObject()
+  , devices_(CallDevices::instance())
 {
         qRegisterMetaType<webrtc::State>();
         qmlRegisterUncreatableMetaObject(
@@ -68,9 +69,7 @@ WebRTCSession::init(std::string *errorMessage)
         gchar *version = gst_version_string();
         nhlog::ui()->info("WebRTC: initialised {}", version);
         g_free(version);
-#if GST_CHECK_VERSION(1, 18, 0)
-        startDeviceMonitor();
-#endif
+        devices_.init();
         return true;
 #else
         (void)errorMessage;
@@ -81,195 +80,17 @@ WebRTCSession::init(std::string *errorMessage)
 #ifdef GSTREAMER_AVAILABLE
 namespace {
 
-struct AudioSource
-{
-        std::string name;
-        GstDevice *device;
-};
-
-struct VideoSource
-{
-        struct Caps
-        {
-                std::string resolution;
-                std::vector<std::string> frameRates;
-        };
-        std::string name;
-        GstDevice *device;
-        std::vector<Caps> caps;
-};
-
 std::string localsdp_;
 std::vector<mtx::events::msg::CallCandidates::Candidate> localcandidates_;
 bool haveAudioStream_;
 bool haveVideoStream_;
-std::vector<AudioSource> audioSources_;
-std::vector<VideoSource> videoSources_;
 GstPad *insetSinkPad_ = nullptr;
 
-using FrameRate = std::pair<int, int>;
-std::optional<FrameRate>
-getFrameRate(const GValue *value)
-{
-        if (GST_VALUE_HOLDS_FRACTION(value)) {
-                gint num = gst_value_get_fraction_numerator(value);
-                gint den = gst_value_get_fraction_denominator(value);
-                return FrameRate{num, den};
-        }
-        return std::nullopt;
-}
-
-void
-addFrameRate(std::vector<std::string> &rates, const FrameRate &rate)
-{
-        constexpr double minimumFrameRate = 15.0;
-        if (static_cast<double>(rate.first) / rate.second >= minimumFrameRate)
-                rates.push_back(std::to_string(rate.first) + "/" + std::to_string(rate.second));
-}
-
-std::pair<int, int>
-tokenise(std::string_view str, char delim)
-{
-        std::pair<int, int> ret;
-        ret.first  = std::atoi(str.data());
-        auto pos   = str.find_first_of(delim);
-        ret.second = std::atoi(str.data() + pos + 1);
-        return ret;
-}
-
-void
-addDevice(GstDevice *device)
-{
-        if (!device)
-                return;
-
-        gchar *name  = gst_device_get_display_name(device);
-        gchar *type  = gst_device_get_device_class(device);
-        bool isVideo = !std::strncmp(type, "Video", 5);
-        g_free(type);
-        nhlog::ui()->debug("WebRTC: {} device added: {}", isVideo ? "video" : "audio", name);
-        if (!isVideo) {
-                audioSources_.push_back({name, device});
-                g_free(name);
-                return;
-        }
-
-        GstCaps *gstcaps = gst_device_get_caps(device);
-        if (!gstcaps) {
-                nhlog::ui()->debug("WebRTC: unable to get caps for {}", name);
-                g_free(name);
-                return;
-        }
-
-        VideoSource source{name, device, {}};
-        g_free(name);
-        guint nCaps = gst_caps_get_size(gstcaps);
-        for (guint i = 0; i < nCaps; ++i) {
-                GstStructure *structure = gst_caps_get_structure(gstcaps, i);
-                const gchar *name       = gst_structure_get_name(structure);
-                if (!std::strcmp(name, "video/x-raw")) {
-                        gint widthpx, heightpx;
-                        if (gst_structure_get(structure,
-                                              "width",
-                                              G_TYPE_INT,
-                                              &widthpx,
-                                              "height",
-                                              G_TYPE_INT,
-                                              &heightpx,
-                                              nullptr)) {
-                                VideoSource::Caps caps;
-                                caps.resolution =
-                                  std::to_string(widthpx) + "x" + std::to_string(heightpx);
-                                const GValue *value =
-                                  gst_structure_get_value(structure, "framerate");
-                                if (auto fr = getFrameRate(value); fr)
-                                        addFrameRate(caps.frameRates, *fr);
-                                else if (GST_VALUE_HOLDS_FRACTION_RANGE(value)) {
-                                        const GValue *minRate =
-                                          gst_value_get_fraction_range_min(value);
-                                        if (auto fr = getFrameRate(minRate); fr)
-                                                addFrameRate(caps.frameRates, *fr);
-                                        const GValue *maxRate =
-                                          gst_value_get_fraction_range_max(value);
-                                        if (auto fr = getFrameRate(maxRate); fr)
-                                                addFrameRate(caps.frameRates, *fr);
-                                } else if (GST_VALUE_HOLDS_LIST(value)) {
-                                        guint nRates = gst_value_list_get_size(value);
-                                        for (guint j = 0; j < nRates; ++j) {
-                                                const GValue *rate =
-                                                  gst_value_list_get_value(value, j);
-                                                if (auto fr = getFrameRate(rate); fr)
-                                                        addFrameRate(caps.frameRates, *fr);
-                                        }
-                                }
-                                if (!caps.frameRates.empty())
-                                        source.caps.push_back(std::move(caps));
-                        }
-                }
-        }
-        gst_caps_unref(gstcaps);
-        videoSources_.push_back(std::move(source));
-}
-
-#if GST_CHECK_VERSION(1, 18, 0)
-template<typename T>
-bool
-removeDevice(T &sources, GstDevice *device, bool changed)
-{
-        if (auto it = std::find_if(sources.begin(),
-                                   sources.end(),
-                                   [device](const auto &s) { return s.device == device; });
-            it != sources.end()) {
-                nhlog::ui()->debug(std::string("WebRTC: device ") +
-                                     (changed ? "changed: " : "removed: ") + "{}",
-                                   it->name);
-                gst_object_unref(device);
-                sources.erase(it);
-                return true;
-        }
-        return false;
-}
-
-void
-removeDevice(GstDevice *device, bool changed)
-{
-        if (device) {
-                if (removeDevice(audioSources_, device, changed) ||
-                    removeDevice(videoSources_, device, changed))
-                        return;
-        }
-}
-#endif
-
 gboolean
 newBusMessage(GstBus *bus G_GNUC_UNUSED, GstMessage *msg, gpointer user_data)
 {
         WebRTCSession *session = static_cast<WebRTCSession *>(user_data);
         switch (GST_MESSAGE_TYPE(msg)) {
-#if GST_CHECK_VERSION(1, 18, 0)
-        case GST_MESSAGE_DEVICE_ADDED: {
-                GstDevice *device;
-                gst_message_parse_device_added(msg, &device);
-                addDevice(device);
-                emit WebRTCSession::instance().devicesChanged();
-                break;
-        }
-        case GST_MESSAGE_DEVICE_REMOVED: {
-                GstDevice *device;
-                gst_message_parse_device_removed(msg, &device);
-                removeDevice(device, false);
-                emit WebRTCSession::instance().devicesChanged();
-                break;
-        }
-        case GST_MESSAGE_DEVICE_CHANGED: {
-                GstDevice *device;
-                GstDevice *oldDevice;
-                gst_message_parse_device_changed(msg, &device, &oldDevice);
-                removeDevice(oldDevice, true);
-                addDevice(device);
-                break;
-        }
-#endif
         case GST_MESSAGE_EOS:
                 nhlog::ui()->error("WebRTC: end of stream");
                 session->end();
@@ -724,27 +545,6 @@ getMediaAttributes(const GstSDPMessage *sdp,
         return false;
 }
 
-template<typename T>
-std::vector<std::string>
-deviceNames(T &sources, const std::string &defaultDevice)
-{
-        std::vector<std::string> ret;
-        ret.reserve(sources.size());
-        std::transform(sources.cbegin(),
-                       sources.cend(),
-                       std::back_inserter(ret),
-                       [](const auto &s) { return s.name; });
-
-        // move default device to top of the list
-        if (auto it = std::find_if(ret.begin(),
-                                   ret.end(),
-                                   [&defaultDevice](const auto &s) { return s == defaultDevice; });
-            it != ret.end())
-                std::swap(ret.front(), *it);
-
-        return ret;
-}
-
 }
 
 bool
@@ -995,19 +795,11 @@ WebRTCSession::startPipeline(int opusPayloadType, int vp8PayloadType)
 bool
 WebRTCSession::createPipeline(int opusPayloadType, int vp8PayloadType)
 {
-        std::string microphoneSetting =
-          ChatPage::instance()->userSettings()->microphone().toStdString();
-        auto it =
-          std::find_if(audioSources_.cbegin(),
-                       audioSources_.cend(),
-                       [&microphoneSetting](const auto &s) { return s.name == microphoneSetting; });
-        if (it == audioSources_.cend()) {
-                nhlog::ui()->error("WebRTC: unknown microphone: {}", microphoneSetting);
+        GstDevice *device = devices_.audioDevice();
+        if (!device)
                 return false;
-        }
-        nhlog::ui()->debug("WebRTC: microphone: {}", microphoneSetting);
 
-        GstElement *source     = gst_device_create_element(it->device, nullptr);
+        GstElement *source     = gst_device_create_element(device, nullptr);
         GstElement *volume     = gst_element_factory_make("volume", "srclevel");
         GstElement *convert    = gst_element_factory_make("audioconvert", nullptr);
         GstElement *resample   = gst_element_factory_make("audioresample", nullptr);
@@ -1070,30 +862,16 @@ bool
 WebRTCSession::addVideoPipeline(int vp8PayloadType)
 {
         // allow incoming video calls despite localUser having no webcam
-        if (videoSources_.empty())
+        if (!devices_.haveCamera())
                 return !isOffering_;
 
-        QSharedPointer<UserSettings> settings = ChatPage::instance()->userSettings();
-        std::string cameraSetting             = settings->camera().toStdString();
-        auto it                               = std::find_if(videoSources_.cbegin(),
-                               videoSources_.cend(),
-                               [&cameraSetting](const auto &s) { return s.name == cameraSetting; });
-        if (it == videoSources_.cend()) {
-                nhlog::ui()->error("WebRTC: unknown camera: {}", cameraSetting);
+        std::pair<int, int> resolution;
+        std::pair<int, int> frameRate;
+        GstDevice *device = devices_.videoDevice(resolution, frameRate);
+        if (!device)
                 return false;
-        }
 
-        std::string resSetting = settings->cameraResolution().toStdString();
-        const std::string &res = resSetting.empty() ? it->caps.front().resolution : resSetting;
-        std::string frSetting  = settings->cameraFrameRate().toStdString();
-        const std::string &fr = frSetting.empty() ? it->caps.front().frameRates.front() : frSetting;
-        auto resolution       = tokenise(res, 'x');
-        auto frameRate        = tokenise(fr, '/');
-        nhlog::ui()->debug("WebRTC: camera: {}", cameraSetting);
-        nhlog::ui()->debug("WebRTC: camera resolution: {}x{}", resolution.first, resolution.second);
-        nhlog::ui()->debug("WebRTC: camera frame rate: {}/{}", frameRate.first, frameRate.second);
-
-        GstElement *source       = gst_device_create_element(it->device, nullptr);
+        GstElement *source       = gst_device_create_element(device, nullptr);
         GstElement *videoconvert = gst_element_factory_make("videoconvert", nullptr);
         GstElement *capsfilter   = gst_element_factory_make("capsfilter", "camerafilter");
         GstCaps *caps            = gst_caps_new_simple("video/x-raw",
@@ -1239,111 +1017,6 @@ WebRTCSession::end()
                 emit stateChanged(State::DISCONNECTED);
 }
 
-#if GST_CHECK_VERSION(1, 18, 0)
-void
-WebRTCSession::startDeviceMonitor()
-{
-        if (!initialised_)
-                return;
-
-        static GstDeviceMonitor *monitor = nullptr;
-        if (!monitor) {
-                monitor       = gst_device_monitor_new();
-                GstCaps *caps = gst_caps_new_empty_simple("audio/x-raw");
-                gst_device_monitor_add_filter(monitor, "Audio/Source", caps);
-                gst_caps_unref(caps);
-                caps = gst_caps_new_empty_simple("video/x-raw");
-                gst_device_monitor_add_filter(monitor, "Video/Source", caps);
-                gst_caps_unref(caps);
-
-                GstBus *bus = gst_device_monitor_get_bus(monitor);
-                gst_bus_add_watch(bus, newBusMessage, nullptr);
-                gst_object_unref(bus);
-                if (!gst_device_monitor_start(monitor)) {
-                        nhlog::ui()->error("WebRTC: failed to start device monitor");
-                        return;
-                }
-        }
-}
-#endif
-
-void
-WebRTCSession::refreshDevices()
-{
-#if GST_CHECK_VERSION(1, 18, 0)
-        return;
-#else
-        if (!initialised_)
-                return;
-
-        static GstDeviceMonitor *monitor = nullptr;
-        if (!monitor) {
-                monitor       = gst_device_monitor_new();
-                GstCaps *caps = gst_caps_new_empty_simple("audio/x-raw");
-                gst_device_monitor_add_filter(monitor, "Audio/Source", caps);
-                gst_caps_unref(caps);
-                caps = gst_caps_new_empty_simple("video/x-raw");
-                gst_device_monitor_add_filter(monitor, "Video/Source", caps);
-                gst_caps_unref(caps);
-        }
-
-        auto clearDevices = [](auto &sources) {
-                std::for_each(
-                  sources.begin(), sources.end(), [](auto &s) { gst_object_unref(s.device); });
-                sources.clear();
-        };
-        clearDevices(audioSources_);
-        clearDevices(videoSources_);
-
-        GList *devices = gst_device_monitor_get_devices(monitor);
-        if (devices) {
-                for (GList *l = devices; l != nullptr; l = l->next)
-                        addDevice(GST_DEVICE_CAST(l->data));
-                g_list_free(devices);
-        }
-        emit devicesChanged();
-#endif
-}
-
-std::vector<std::string>
-WebRTCSession::getDeviceNames(bool isVideo, const std::string &defaultDevice) const
-{
-        return isVideo ? deviceNames(videoSources_, defaultDevice)
-                       : deviceNames(audioSources_, defaultDevice);
-}
-
-std::vector<std::string>
-WebRTCSession::getResolutions(const std::string &cameraName) const
-{
-        std::vector<std::string> ret;
-        if (auto it = std::find_if(videoSources_.cbegin(),
-                                   videoSources_.cend(),
-                                   [&cameraName](const auto &s) { return s.name == cameraName; });
-            it != videoSources_.cend()) {
-                ret.reserve(it->caps.size());
-                for (const auto &c : it->caps)
-                        ret.push_back(c.resolution);
-        }
-        return ret;
-}
-
-std::vector<std::string>
-WebRTCSession::getFrameRates(const std::string &cameraName, const std::string &resolution) const
-{
-        if (auto i = std::find_if(videoSources_.cbegin(),
-                                  videoSources_.cend(),
-                                  [&](const auto &s) { return s.name == cameraName; });
-            i != videoSources_.cend()) {
-                if (auto j =
-                      std::find_if(i->caps.cbegin(),
-                                   i->caps.cend(),
-                                   [&](const auto &s) { return s.resolution == resolution; });
-                    j != i->caps.cend())
-                        return j->frameRates;
-        }
-        return {};
-}
-
 #else
 
 bool
@@ -1400,25 +1073,4 @@ void
 WebRTCSession::end()
 {}
 
-void
-WebRTCSession::refreshDevices()
-{}
-
-std::vector<std::string>
-WebRTCSession::getDeviceNames(bool, const std::string &) const
-{
-        return {};
-}
-
-std::vector<std::string>
-WebRTCSession::getResolutions(const std::string &) const
-{
-        return {};
-}
-
-std::vector<std::string>
-WebRTCSession::getFrameRates(const std::string &, const std::string &) const
-{
-        return {};
-}
 #endif
diff --git a/src/WebRTCSession.h b/src/WebRTCSession.h
index 2f0fb70e..0fe8a864 100644
--- a/src/WebRTCSession.h
+++ b/src/WebRTCSession.h
@@ -5,6 +5,7 @@
 
 #include <QObject>
 
+#include "CallDevices.h"
 #include "mtx/events/voip.hpp"
 
 typedef struct _GstElement GstElement;
@@ -59,13 +60,6 @@ public:
 
         void setTurnServers(const std::vector<std::string> &uris) { turnServers_ = uris; }
 
-        void refreshDevices();
-        std::vector<std::string> getDeviceNames(bool isVideo,
-                                                const std::string &defaultDevice) const;
-        std::vector<std::string> getResolutions(const std::string &cameraName) const;
-        std::vector<std::string> getFrameRates(const std::string &cameraName,
-                                               const std::string &resolution) const;
-
         void setVideoItem(QQuickItem *item) { videoItem_ = item; }
         QQuickItem *getVideoItem() const { return videoItem_; }
 
@@ -76,7 +70,6 @@ signals:
                            const std::vector<mtx::events::msg::CallCandidates::Candidate> &);
         void newICECandidate(const mtx::events::msg::CallCandidates::Candidate &);
         void stateChanged(webrtc::State);
-        void devicesChanged();
 
 private slots:
         void setState(webrtc::State state) { state_ = state; }
@@ -84,6 +77,7 @@ private slots:
 private:
         WebRTCSession();
 
+        CallDevices &devices_;
         bool initialised_           = false;
         bool haveVoicePlugins_      = false;
         bool haveVideoPlugins_      = false;
@@ -101,7 +95,6 @@ private:
         bool startPipeline(int opusPayloadType, int vp8PayloadType);
         bool createPipeline(int opusPayloadType, int vp8PayloadType);
         bool addVideoPipeline(int vp8PayloadType);
-        void startDeviceMonitor();
 
 public:
         WebRTCSession(WebRTCSession const &) = delete;