basic voice capture + transmission

4 files changed, 132 insertions, 8 deletions

diff --git a/src/audio/manager.cpp b/src/audio/manager.cpp
index 9d380f9..b665c81 100644
--- a/src/audio/manager.cpp
+++ b/src/audio/manager.cpp
@@ -39,9 +39,25 @@ void data_callback(ma_device *pDevice, void *pOutput, const void *pInput, ma_uin
     }
 }
 
+void capture_data_callback(ma_device *pDevice, void *pOutput, const void *pInput, ma_uint32 frameCount) {
+    auto *mgr = reinterpret_cast<AudioManager *>(pDevice->pUserData);
+    if (mgr == nullptr) return;
+
+    mgr->OnCapturedPCM(static_cast<const int16_t *>(pInput), frameCount);
+}
+
 AudioManager::AudioManager() {
     m_ok = true;
 
+    int err;
+    m_encoder = opus_encoder_create(48000, 2, OPUS_APPLICATION_VOIP, &err);
+    if (err != OPUS_OK) {
+        printf("failed to initialize opus encoder: %d\n", err);
+        m_ok = false;
+        return;
+    }
+    opus_encoder_ctl(m_encoder, OPUS_SET_BITRATE(64000));
+
     m_device_config = ma_device_config_init(ma_device_type_playback);
     m_device_config.playback.format = ma_format_f32;
     m_device_config.playback.channels = 2;
@@ -50,7 +66,7 @@ AudioManager::AudioManager() {
     m_device_config.pUserData = this;
 
     if (ma_device_init(nullptr, &m_device_config, &m_device) != MA_SUCCESS) {
-        puts("open playabck fail");
+        puts("open playback fail");
         m_ok = false;
         return;
     }
@@ -61,15 +77,45 @@ AudioManager::AudioManager() {
         m_ok = false;
         return;
     }
+
+    m_capture_config = ma_device_config_init(ma_device_type_capture);
+    m_capture_config.capture.format = ma_format_s16;
+    m_capture_config.capture.channels = 2;
+    m_capture_config.sampleRate = 48000;
+    m_capture_config.periodSizeInFrames = 480;
+    m_capture_config.dataCallback = capture_data_callback;
+    m_capture_config.pUserData = this;
+
+    if (ma_device_init(nullptr, &m_capture_config, &m_capture_device) != MA_SUCCESS) {
+        puts("open capture fail");
+        m_ok = false;
+        return;
+    }
+
+    if (ma_device_start(&m_capture_device) != MA_SUCCESS) {
+        puts("failed to start capture");
+        ma_device_uninit(&m_capture_device);
+        m_ok = false;
+        return;
+    }
+
+    char device_name[MA_MAX_DEVICE_NAME_LENGTH + 1];
+    ma_device_get_name(&m_capture_device, ma_device_type_capture, device_name, sizeof(device_name), nullptr);
+    printf("using %s for capture\n", device_name);
 }
 
 AudioManager::~AudioManager() {
     ma_device_uninit(&m_device);
+    ma_device_uninit(&m_capture_device);
     for (auto &[ssrc, pair] : m_sources) {
         opus_decoder_destroy(pair.second);
     }
 }
 
+void AudioManager::SetOpusBuffer(uint8_t *ptr) {
+    m_opus_buffer = ptr;
+}
+
 void AudioManager::FeedMeOpus(uint32_t ssrc, const std::vector<uint8_t> &data) {
     size_t payload_size = 0;
     const auto *opus_encoded = StripRTPExtensionHeader(data.data(), static_cast<int>(data.size()), payload_size);
@@ -89,7 +135,22 @@ void AudioManager::FeedMeOpus(uint32_t ssrc, const std::vector<uint8_t> &data) {
     }
 }
 
+void AudioManager::OnCapturedPCM(const int16_t *pcm, ma_uint32 frames) {
+    if (m_opus_buffer == nullptr) return;
+
+    int payload_len = opus_encode(m_encoder, pcm, 480, static_cast<unsigned char *>(m_opus_buffer), 1275);
+    if (payload_len < 0) {
+        printf("encoding error: %d\n", payload_len);
+    } else {
+        m_signal_opus_packet.emit(payload_len);
+    }
+}
+
 bool AudioManager::OK() const {
     return m_ok;
 }
+
+AudioManager::type_signal_opus_packet AudioManager::signal_opus_packet() {
+    return m_signal_opus_packet;
+}
 #endif
diff --git a/src/audio/manager.hpp b/src/audio/manager.hpp
index d0f3a21..700fcc0 100644
--- a/src/audio/manager.hpp
+++ b/src/audio/manager.hpp
@@ -10,6 +10,7 @@
 #include <vector>
 #include <miniaudio.h>
 #include <opus.h>
+#include <sigc++/sigc++.h>
 // clang-format on
 
 class AudioManager {
@@ -17,21 +18,40 @@ public:
     AudioManager();
     ~AudioManager();
 
+    void SetOpusBuffer(uint8_t *ptr);
     void FeedMeOpus(uint32_t ssrc, const std::vector<uint8_t> &data);
 
     [[nodiscard]] bool OK() const;
 
 private:
+    void OnCapturedPCM(const int16_t *pcm, ma_uint32 frames);
+
     friend void data_callback(ma_device *, void *, const void *, ma_uint32);
+    friend void capture_data_callback(ma_device *, void *, const void *, ma_uint32);
 
     std::thread m_thread;
 
     bool m_ok;
 
+    // playback
     ma_device m_device;
     ma_device_config m_device_config;
+    // capture
+    ma_device m_capture_device;
+    ma_device_config m_capture_config;
 
     std::mutex m_mutex;
     std::unordered_map<uint32_t, std::pair<std::deque<int16_t>, OpusDecoder *>> m_sources;
+
+    OpusEncoder *m_encoder;
+
+    uint8_t *m_opus_buffer = nullptr;
+
+public:
+    using type_signal_opus_packet = sigc::signal<void(int payload_size)>;
+    type_signal_opus_packet signal_opus_packet();
+
+private:
+    type_signal_opus_packet m_signal_opus_packet;
 };
 #endif
diff --git a/src/discord/voiceclient.cpp b/src/discord/voiceclient.cpp
index d8855fd..3f38eea 100644
--- a/src/discord/voiceclient.cpp
+++ b/src/discord/voiceclient.cpp
@@ -43,11 +43,11 @@ void UDPSocket::SetSSRC(uint32_t ssrc) {
     m_ssrc = ssrc;
 }
 
-void UDPSocket::SendEncrypted(const std::vector<uint8_t> &data) {
+void UDPSocket::SendEncrypted(const uint8_t *data, size_t len) {
     m_sequence++;
-    m_timestamp += (48000 / 100) * 2;
+    m_timestamp += 480; // this is important
 
-    std::vector<uint8_t> rtp(12, 0);
+    std::vector<uint8_t> rtp(12 + len + crypto_secretbox_MACBYTES, 0);
     rtp[0] = 0x80; // ver 2
     rtp[1] = 0x78; // payload type 0x78
     rtp[2] = (m_sequence >> 8) & 0xFF;
@@ -63,14 +63,15 @@ void UDPSocket::SendEncrypted(const std::vector<uint8_t> &data) {
 
     static std::array<uint8_t, 24> nonce = {};
     std::memcpy(nonce.data(), rtp.data(), 12);
-
-    std::vector<uint8_t> ciphertext(crypto_secretbox_MACBYTES + rtp.size(), 0);
-    crypto_secretbox_easy(ciphertext.data(), rtp.data(), rtp.size(), nonce.data(), m_secret_key.data());
-    rtp.insert(rtp.end(), ciphertext.begin(), ciphertext.end());
+    crypto_secretbox_easy(rtp.data() + 12, data, len, nonce.data(), m_secret_key.data());
 
     Send(rtp.data(), rtp.size());
 }
 
+void UDPSocket::SendEncrypted(const std::vector<uint8_t> &data) {
+    SendEncrypted(data.data(), data.size());
+}
+
 void UDPSocket::Send(const uint8_t *data, size_t len) {
     sendto(m_socket, reinterpret_cast<const char *>(data), static_cast<int>(len), 0, reinterpret_cast<sockaddr *>(&m_server), sizeof(m_server));
 }
@@ -172,6 +173,14 @@ DiscordVoiceClient::~DiscordVoiceClient() {
 
 void DiscordVoiceClient::Start() {
     m_ws.StartConnection("wss://" + m_endpoint + "/?v=7");
+
+    // cant put in ctor or deadlock in singleton initialization
+    auto &aud = Abaddon::Get().GetAudio();
+    aud.SetOpusBuffer(m_opus_buffer.data());
+    aud.signal_opus_packet().connect([this](int payload_size) {
+        if (m_connected)
+            m_udp.SendEncrypted(m_opus_buffer.data(), payload_size);
+    });
 }
 
 void DiscordVoiceClient::SetSessionID(std::string_view session_id) {
@@ -241,6 +250,13 @@ void DiscordVoiceClient::HandleGatewaySessionDescription(const VoiceGatewayMessa
         printf("%02X", b);
     }
     printf("\n");
+
+    VoiceSpeakingMessage msg;
+    msg.Delay = 0;
+    msg.SSRC = m_ssrc;
+    msg.Speaking = VoiceSpeakingMessage::Microphone;
+    m_ws.Send(msg);
+
     m_secret_key = d.SecretKey;
     m_udp.SetSSRC(m_ssrc);
     m_udp.SetSecretKey(m_secret_key);
@@ -250,6 +266,7 @@ void DiscordVoiceClient::HandleGatewaySessionDescription(const VoiceGatewayMessa
     m_udp.SendEncrypted({ 0xF8, 0xFF, 0xFE });
     m_udp.SendEncrypted({ 0xF8, 0xFF, 0xFE });
     m_udp.Run();
+    m_connected = true;
 }
 
 void DiscordVoiceClient::Identify() {
@@ -388,4 +405,11 @@ void from_json(const nlohmann::json &j, VoiceSessionDescriptionData &m) {
     JS_D("mode", m.Mode);
     JS_D("secret_key", m.SecretKey);
 }
+
+void to_json(nlohmann::json &j, const VoiceSpeakingMessage &m) {
+    j["op"] = VoiceGatewayOp::Speaking;
+    j["d"]["speaking"] = m.Speaking;
+    j["d"]["delay"] = m.Delay;
+    j["d"]["ssrc"] = m.SSRC;
+}
 #endif
diff --git a/src/discord/voiceclient.hpp b/src/discord/voiceclient.hpp
index 4b988d5..67919e5 100644
--- a/src/discord/voiceclient.hpp
+++ b/src/discord/voiceclient.hpp
@@ -110,6 +110,20 @@ struct VoiceSessionDescriptionData {
     friend void from_json(const nlohmann::json &j, VoiceSessionDescriptionData &m);
 };
 
+struct VoiceSpeakingMessage {
+    enum {
+        Microphone = 1 << 0,
+        Soundshare = 1 << 1,
+        Priority = 1 << 2,
+    };
+
+    int Speaking;
+    int Delay;
+    uint32_t SSRC;
+
+    friend void to_json(nlohmann::json &j, const VoiceSpeakingMessage &m);
+};
+
 class UDPSocket {
 public:
     UDPSocket();
@@ -119,6 +133,7 @@ public:
     void Run();
     void SetSecretKey(std::array<uint8_t, 32> key);
     void SetSSRC(uint32_t ssrc);
+    void SendEncrypted(const uint8_t *data, size_t len);
     void SendEncrypted(const std::vector<uint8_t> &data);
     void Send(const uint8_t *data, size_t len);
     std::vector<uint8_t> Receive();
@@ -205,5 +220,9 @@ private:
     int m_heartbeat_msec;
     Waiter m_heartbeat_waiter;
     std::thread m_heartbeat_thread;
+
+    std::array<uint8_t, 1275> m_opus_buffer;
+
+    std::atomic<bool> m_connected = false;
 };
 #endif