feat(audio): implement voice chat using Opus codec

Add voice chat support with Opus encoding/decoding, push-to-talk (V key), and new audio recording infrastructure.

CMakeLists.txt

@@ -147,6 +147,7 @@ target_link_libraries(${PROJECT_NAME}
         zstd::zstd
         OpenAL::OpenAL
         harfbuzz::harfbuzz
+        Opus::Opus
         $<$<PLATFORM_ID:Windows>:ws2_32>
 
 )

cmake/Dependencies.cmake

@@ -9,7 +9,7 @@ find_package(OpenAL REQUIRED)
 find_package(harfbuzz REQUIRED)
 find_package(Freetype REQUIRED)
 find_package(SDL3 REQUIRED)
-
+find_package(Opus REQUIRED)
 
 # Third-party libraries
 FetchContent_Declare(

cmake/modules/FindOpus.cmake

@@ -0,0 +1,25 @@
+include(FindPackageHandleStandardArgs)
+
+find_path(OPUS_INCLUDE_DIR
+    NAMES opus/opus.h opus.h
+)
+
+find_library(OPUS_LIBRARY
+    NAMES opus
+)
+
+find_package_handle_standard_args(
+    Opus
+    REQUIRED_VARS
+        OPUS_INCLUDE_DIR
+        OPUS_LIBRARY
+)
+
+if(Opus_FOUND)
+    add_library(Opus::Opus UNKNOWN IMPORTED)
+
+    set_target_properties(Opus::Opus PROPERTIES
+        IMPORTED_LOCATION "${OPUS_LIBRARY}"
+        INTERFACE_INCLUDE_DIRECTORIES "${OPUS_INCLUDE_DIR}"
+    )
+endif()

include/Cubed/audio/audio_engine.hpp

@@ -1,6 +1,7 @@
 #pragma once
 
 #include "Cubed/audio/audio_fade.hpp"
+#include "Cubed/audio/audio_recording.hpp"
 #include "Cubed/audio/audio_source.hpp"
 #include "Cubed/audio/sound_manager.hpp"
 #include "Cubed/audio/source_pool.hpp"
@@ -10,13 +11,15 @@
 #include <AL/alc.h>
 #include <glm/glm.hpp>
 #include <memory>
+#include <opus/opus.h>
 #include <string>
 #include <unordered_map>
 namespace Cubed {
 
-class ClientWorld;
+class NetworkClient;
 
 class AudioEngine {
+
 public:
     AudioEngine(Config& config);
     AudioEngine(const AudioEngine&) = delete;
@@ -41,11 +44,22 @@ public:
 
     float& bgm_target_volume();
 
+    void set_client(std::weak_ptr<NetworkClient> client);
+
+    void send_voice(const std::array<int16_t, AudioRecording::FRAME_SAMPLES>&);
+    void receive_voice(std::span<char> opus, const glm::vec3& pos);
+
+    AudioRecording& audio_recording();
+
 private:
     using FadeMap = std::unordered_map<std::string, AudioFade>;
     bool m_init{false};
     ALCdevice* device{nullptr};
     ALCcontext* context{nullptr};
+    OpusEncoder* m_encoder{nullptr};
+    OpusDecoder* m_decoder{nullptr};
+    AudioRecording m_recording;
+    std::weak_ptr<NetworkClient> m_client;
     glm::vec3 listener_pos;
     std::unique_ptr<AudioSource> m_bgm;
     FadeMap m_fade_map;

include/Cubed/audio/audio_recording.hpp

@@ -0,0 +1,33 @@
+#pragma once
+
+#include <alc.h>
+#include <array>
+#include <cstdint>
+namespace Cubed {
+class AudioEngine;
+class AudioRecording {
+public:
+    static constexpr int SAMPLE_RATE = 48000;
+    static constexpr int FRAME_MS = 20;
+    static constexpr int FRAME_SAMPLES = SAMPLE_RATE * FRAME_MS / 1000;
+    AudioRecording(AudioEngine& engine);
+    AudioRecording(const AudioRecording&) = delete;
+    AudioRecording(AudioRecording&&) = delete;
+    AudioRecording& operator=(const AudioRecording&) = delete;
+    AudioRecording& operator=(AudioRecording&&) = delete;
+    ~AudioRecording();
+
+    void update();
+
+    void init();
+    void start();
+    void stop();
+    bool is_recording() const;
+
+private:
+    AudioEngine& m_engine;
+    ALCdevice* m_capture = nullptr;
+    bool m_recording = false;
+    void send_voice(const std::array<int16_t, FRAME_SAMPLES>& pcm);
+};
+} // namespace Cubed

include/Cubed/audio/audio_source.hpp

@@ -5,6 +5,7 @@
 
 #include <AL/al.h>
 #include <glm/glm.hpp>
+#include <memory>
 namespace Cubed {
 
 enum class AudioState { INITIAL, PLAYING, PAUSED, STOPPED };
@@ -21,8 +22,9 @@ public:
     void set_pitch(float pitch);
     void play();
     void play_2d(const AudioBuffer& buffer);
+    void play_2d(std::unique_ptr<AudioBuffer> buffer);
     void play_3d(const AudioBuffer& buffer, const glm::vec3& pos);
-
+    void play_3d(std::unique_ptr<AudioBuffer> buffer, const glm::vec3& pos);
     void stop();
     void pause();
     float duration() const;
@@ -42,6 +44,8 @@ public:
     void clear_effect_slot();
 
 private:
+    std::unique_ptr<AudioBuffer> m_buffer;
+
     ALuint m_source = 0;
     float m_target_volume = 1.0f;
     float m_duration = 0.0f;

include/Cubed/audio/source_pool.hpp

@@ -4,7 +4,7 @@
 namespace Cubed {
 class SourcePool {
 private:
-    std::vector<AudioSource> m_sources;
+    std::vector<std::unique_ptr<AudioSource>> m_sources;
 
 public:
     explicit SourcePool(size_t size);
@@ -15,9 +15,10 @@ public:
     SourcePool& operator=(SourcePool&&) = delete;
 
     void update();
+    [[nodiscard]]
     AudioSource* acquire();
 
-    std::vector<AudioSource>& sources();
+    std::vector<std::unique_ptr<AudioSource>>& sources();
 };
 
 } // namespace Cubed

include/Cubed/gameplay/client_world.hpp

@@ -106,7 +106,7 @@ public:
 
     void receive_chat_message(ChatMsg& msg);
     void send_chat_message(ChatMessage& message);
-
+    void receive_voice_message(VoiceMsg& msg);
     template <typename Fn>
     void register_ticktimer(std::string_view id, TickType threshold, Fn&& f) {
         m_ticktimers.emplace(
@@ -115,6 +115,11 @@ public:
     }
 
 private:
+    struct VoiceMessage {
+        std::string data;
+        glm::vec3 pos;
+    };
+
     std::atomic<bool> m_is_pending_delete_queue_free{false};
     std::mutex m_delete_vbo_mutex;
     std::mutex m_delete_vao_mutex;
@@ -153,6 +158,7 @@ private:
     tbb::concurrent_queue<ChunkPos> m_dirty_chunk_queue;
     tbb::concurrent_queue<PendingSound> m_pending_sound;
     tbb::concurrent_queue<ChatMessage> m_message_queue;
+    tbb::concurrent_queue<VoiceMessage> m_voice_queue;
 
     std::deque<ChunkPos> m_dirty_queue;
     std::vector<const ChunkRenderSnapshot*> m_render_snapshots;

include/Cubed/gameplay/network_client.hpp

@@ -21,6 +21,7 @@ public:
     bool is_connect_error() const;
     std::string get_error_string() const;
     void clear_error();
+    ClientWorld& world();
 
 private:
     struct Task {

include/Cubed/gameplay/packet.hpp

@@ -62,6 +62,7 @@ enum class PacketEnum : uint16_t {
     S2C_CLEAR_ALL_CHUNKS = 3005,
     UPDATE_TIME = 3006,
     CHAT_MSG = 4001,
+    VOICE_MSG = 4002,
     PING = 9001,
     PONG = 9002
 
@@ -125,6 +126,9 @@ template <> constexpr uint16_t get_packet_id<PlayerWaterSound>() {
 template <> constexpr uint16_t get_packet_id<ChatMsg>() {
     return std::to_underlying(PacketEnum::CHAT_MSG);
 }
+template <> constexpr uint16_t get_packet_id<VoiceMsg>() {
+    return std::to_underlying(PacketEnum::VOICE_MSG);
+}
 
 template <typename T>
     requires std::derived_from<T, google::protobuf::Message>

include/Cubed/gameplay/server_world.hpp

@@ -81,7 +81,7 @@ public:
     void handle_block_change(const BlockChangeReq& req);
 
     void handle_chat_message(ChatMsg& msg);
-
+    void handle_voice_message(VoiceMsg& msg);
     int chunk_size() const;
     template <typename Fn>
     void register_timer(std::string_view id, TickType threshold, Fn&& f) {

src/CMakeLists.txt

@@ -86,4 +86,5 @@ target_sources(${PROJECT_NAME}
         localization.cpp
         tools/system_locate.cpp
         ui/chat_box.cpp
+        audio/audio_recording.cpp
 )

src/audio/audio_engine.cpp

@@ -1,13 +1,19 @@
 #include "Cubed/audio/audio_engine.hpp"
 
 #include "Cubed/audio/audio_error.hpp"
+#include "Cubed/gameplay/client_world.hpp"
+#include "Cubed/gameplay/network_client.hpp"
 #include "Cubed/tools/cubed_assert.hpp"
 #include "Cubed/tools/log.hpp"
 
 #include <stdexcept>
-
+using namespace google::protobuf;
+namespace {
+constexpr std::size_t OPUS_MAX_PACKET_SIZE = 400;
+}
 namespace Cubed {
-AudioEngine::AudioEngine(Config& config) : m_config(config) {};
+AudioEngine::AudioEngine(Config& config)
+    : m_recording(*this), m_config(config) {};
 
 AudioEngine::~AudioEngine() {
     if (!m_init) {
@@ -21,13 +27,35 @@ AudioEngine::~AudioEngine() {
     m_low_pass_filter.reset();
     m_underwater_effect.reset();
     m_underwater_slot.reset();
-
+    opus_encoder_destroy(m_encoder);
+    opus_decoder_destroy(m_decoder);
     alcMakeContextCurrent(nullptr);
     alcDestroyContext(context);
     alcCloseDevice(device);
 }
 
 void AudioEngine::init() {
+    {
+        int error;
+        m_encoder = opus_encoder_create(AudioRecording::SAMPLE_RATE,
+                                        1, // Mono
+                                        OPUS_APPLICATION_VOIP, &error);
+
+        if (error != OPUS_OK) {
+            Logger::error("Can't Create Opus Encoder Error {}", error);
+        }
+        opus_encoder_ctl(m_encoder, OPUS_SET_BITRATE(24000));
+        opus_encoder_ctl(m_encoder, OPUS_SET_COMPLEXITY(5));
+        opus_encoder_ctl(m_encoder, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE));
+    }
+    {
+        int error;
+        m_decoder = opus_decoder_create(AudioRecording::SAMPLE_RATE, 1, &error);
+
+        if (error != OPUS_OK) {
+            Logger::error("Can't Create Opus Encoder Error {}", error);
+        }
+    }
     device = alcOpenDevice(NULL);
     if (!device) {
         throw std::runtime_error("Failed to open OpenAL device.");
@@ -62,7 +90,7 @@ void AudioEngine::init() {
 
     alDistanceModel(AL_INVERSE_DISTANCE_CLAMPED);
     check_al_error();
-
+    m_recording.init();
     m_music_volume = m_config.get("volume.music", 1.0f);
     m_sfx_volume = m_config.get("volume.SFX", 1.0f);
 
@@ -144,10 +172,10 @@ void AudioEngine::play_2d(const std::string& sound, bool check) {
         return;
     }
     auto* source = m_pool->acquire();
-    source->set_volume(m_sfx_volume);
     if (!source) {
         Logger::error("Source is Full");
     }
+    source->set_volume(m_sfx_volume);
 
     try {
         auto& buffer = m_sounds.get_buffer(sound);
@@ -181,6 +209,7 @@ void AudioEngine::update() {
         fade.update();
     }
     m_pool->update();
+    m_recording.update();
 }
 
 void AudioEngine::reload_config() {
@@ -201,11 +230,11 @@ void AudioEngine::underwater_change(bool underwater) {
     }
     for (auto& source : m_pool->sources()) {
         if (m_underwater) {
-            source.set_filter(*m_low_pass_filter);
+            source->set_filter(*m_low_pass_filter);
-            source.set_effect_slot(*m_underwater_slot);
+            source->set_effect_slot(*m_underwater_slot);
         } else {
-            source.clear_filter();
+            source->clear_filter();
-            source.clear_effect_slot();
+            source->clear_effect_slot();
         }
     }
     if (underwater) {
@@ -219,4 +248,87 @@ void AudioEngine::underwater_change(bool underwater) {
 
 float& AudioEngine::bgm_target_volume() { return m_bgm->target_volume(); }
 
+void AudioEngine::set_client(std::weak_ptr<NetworkClient> client) {
+    m_client = client;
+}
+
+void AudioEngine::send_voice(
+    const std::array<int16_t, AudioRecording::FRAME_SAMPLES>& pcm) {
+    {
+        AudioData data;
+        data.pcm = {pcm.begin(), pcm.end()};
+        data.channels = 1;
+        data.sample_rate = AudioRecording::SAMPLE_RATE;
+        auto* source = m_pool->acquire();
+        source->set_volume(m_sfx_volume);
+        if (!source) {
+            Logger::error("Source is Full");
+            return;
+        }
+        std::unique_ptr<AudioBuffer> buffer =
+            std::make_unique<AudioBuffer>(data);
+
+        if (m_efx_supported && m_underwater) {
+            source->set_filter(*m_low_pass_filter);
+            source->set_effect_slot(*m_underwater_slot);
+        }
+        source->play_2d(std::move(buffer));
+    }
+    /*
+    std::array<uint8_t, OPUS_MAX_PACKET_SIZE> opus;
+    int len = opus_encode(m_encoder, pcm.data(), AudioRecording::FRAME_SAMPLES,
+                          opus.data(), opus.size());
+
+    if (len < 0) {
+        Logger::error("Opus encode failed: {}", opus_strerror(len));
+        return;
+    }
+    Logger::info("opus encode len={}", len);
+    if (auto c = m_client.lock()) {
+        Arena arena;
+        auto msg = Arena::Create<VoiceMsg>(&arena);
+        msg->set_uuid(c->world().get_player().get_uuid());
+        msg->set_opus_data(reinterpret_cast<char*>(opus.data()), len);
+        auto* pos = msg->mutable_pos();
+        auto p = c->world().get_player().get_player_pos();
+        pos->set_x(p.x);
+        pos->set_y(p.y);
+        pos->set_z(p.z);
+        c->send(make_packet(*msg));
+    }
+        */
+}
+void AudioEngine::receive_voice(std::span<char> opus, const glm::vec3& pos) {
+    AudioData data;
+    data.pcm.resize(AudioRecording::FRAME_SAMPLES);
+    data.channels = 1;
+    data.sample_rate = AudioRecording::SAMPLE_RATE;
+    int len = opus_decode(
+        m_decoder, reinterpret_cast<const uint8_t*>(opus.data()), opus.size(),
+        data.pcm.data(), AudioRecording::FRAME_SAMPLES, 0);
+    if (len < 0) {
+        Logger::error("Opus decode failed: {}", opus_strerror(len));
+        return;
+    }
+    Logger::info("decode samples={}", len);
+    Logger::info("Receive Vocie and start play");
+    auto* source = m_pool->acquire();
+    source->set_volume(m_sfx_volume);
+    if (!source) {
+        Logger::error("Source is Full");
+        return;
+    }
+    for (int i = 0; i < 10; ++i)
+        Logger::info("recv {}", data.pcm[i]);
+    std::unique_ptr<AudioBuffer> buffer = std::make_unique<AudioBuffer>(data);
+
+    if (m_efx_supported && m_underwater) {
+        source->set_filter(*m_low_pass_filter);
+        source->set_effect_slot(*m_underwater_slot);
+    }
+    source->play_3d(std::move(buffer), pos);
+}
+
+AudioRecording& AudioEngine::audio_recording() { return m_recording; }
+
 } // namespace Cubed

src/audio/audio_recording.cpp

@@ -0,0 +1,74 @@
+#include "Cubed/audio/audio_recording.hpp"
+
+#include "Cubed/audio/audio_engine.hpp"
+#include "Cubed/tools/log.hpp"
+
+#include <al.h>
+
+namespace Cubed {
+AudioRecording::AudioRecording(AudioEngine& engine)
+    : m_engine(engine) {
+
+      };
+AudioRecording::~AudioRecording() {
+    if (m_capture) {
+        if (m_recording) {
+            stop();
+        }
+
+        alcCaptureCloseDevice(m_capture);
+    }
+}
+
+void AudioRecording::init() {
+    m_capture =
+        alcCaptureOpenDevice(nullptr, SAMPLE_RATE, AL_FORMAT_MONO16, 4096);
+    if (m_capture) {
+        Logger::info("Open Audio Capture Success");
+    } else {
+        Logger::error("Fail to Open Audio Capture Device");
+    }
+    ALCint freq = 0;
+    alcGetIntegerv(m_capture, ALC_FREQUENCY, 1, &freq);
+
+    Logger::info("Capture frequency={}", freq);
+}
+
+void AudioRecording::start() {
+    ALCint samples;
+
+    alcGetIntegerv(m_capture, ALC_CAPTURE_SAMPLES, 1, &samples);
+
+    std::vector<int16_t> dump(samples);
+    alcCaptureSamples(m_capture, dump.data(), samples);
+    alcCaptureStart(m_capture);
+    m_recording = true;
+}
+
+void AudioRecording::stop() {
+    alcCaptureStop(m_capture);
+    m_recording = false;
+}
+bool AudioRecording::is_recording() const { return m_recording; }
+void AudioRecording::update() {
+    if (!m_recording || !m_capture) {
+        return;
+    }
+
+    ALCint available = 0;
+    alcGetIntegerv(m_capture, ALC_CAPTURE_SAMPLES, 1, &available);
+    while (available >= FRAME_SAMPLES) {
+        std::array<int16_t, FRAME_SAMPLES> pcm;
+        alcCaptureSamples(m_capture, pcm.data(), FRAME_SAMPLES);
+        for (int i = 0; i < 20; i++) {
+            Logger::info("pcm[{}]={}", i, pcm[i]);
+        }
+        send_voice(pcm);
+        available -= FRAME_SAMPLES;
+    }
+}
+void AudioRecording::send_voice(const std::array<int16_t, FRAME_SAMPLES>& pcm) {
+
+    m_engine.send_voice(pcm);
+}
+} // namespace Cubed

src/audio/audio_source.cpp

@@ -91,12 +91,19 @@ void AudioSource::play_2d(const AudioBuffer& buffer) {
     set_buffer_2d(buffer);
     play();
 }
-
+void AudioSource::play_2d(std::unique_ptr<AudioBuffer> buffer) {
+    m_buffer = std::move(buffer);
+    play_2d(*m_buffer);
+}
 void AudioSource::play_3d(const AudioBuffer& buffer, const glm::vec3& pos) {
     set_buffer_3d(buffer, pos);
     play();
 }
-
+void AudioSource::play_3d(std::unique_ptr<AudioBuffer> buffer,
+                          const glm::vec3& pos) {
+    m_buffer = std::move(buffer);
+    play_3d(*m_buffer, pos);
+}
 void AudioSource::stop() { alSourceStop(m_source); }
 void AudioSource::pause() { alSourcePause(m_source); }
 
@@ -163,7 +170,7 @@ void AudioSource::reset() {
 
     clear_filter();
     clear_effect_slot();
-
+    m_buffer.reset();
     m_duration = 0.0f;
     m_target_volume = 1.0f;
     m_using = false;

src/audio/source_pool.cpp

@@ -1,24 +1,32 @@
 #include "Cubed/audio/source_pool.hpp"
 
 namespace Cubed {
-SourcePool::SourcePool(size_t size) { m_sources.resize(size); }
+SourcePool::SourcePool(size_t size) {
+    m_sources.reserve(size);
+    for (size_t i = 0; i < size; ++i) {
+        m_sources.emplace_back(std::make_unique<AudioSource>());
+    }
+}
 SourcePool::~SourcePool() { m_sources.clear(); }
 void SourcePool::update() {
-    for (auto& source : m_sources) {
+    for (auto& source_ptr : m_sources) {
-        if (source.in_use() && source.state() == AudioState::STOPPED) {
+        if (source_ptr->in_use() &&
-            source.reset();
+            source_ptr->state() == AudioState::STOPPED) {
+            source_ptr->reset();
         }
     }
 }
 
 AudioSource* SourcePool::acquire() {
-    for (auto& source : m_sources) {
+    for (auto& source_ptr : m_sources) {
-        if (!source.in_use()) {
+        if (!source_ptr->in_use()) {
-            source.mark_in_use();
+            source_ptr->mark_in_use();
-            return &source;
+            return source_ptr.get();
         }
     }
     return nullptr;
 }
-std::vector<AudioSource>& SourcePool::sources() { return m_sources; }
+std::vector<std::unique_ptr<AudioSource>>& SourcePool::sources() {
+    return m_sources;
+}
 } // namespace Cubed

src/gameplay/client_world.cpp

@@ -754,6 +754,13 @@ void ClientWorld::request_exit() {
 void ClientWorld::receive_chat_message(ChatMsg& msg) {
     m_message_queue.emplace(msg.name(), msg.msg(), Tools::get_time_ticks());
 }
+
+void ClientWorld::receive_voice_message(VoiceMsg& msg) {
+    glm::vec3 pos{msg.pos().x(), msg.pos().y(), msg.pos().z()};
+    Logger::info("Receive Voice From net");
+    m_voice_queue.emplace(msg.opus_data(), pos);
+}
+
 void ClientWorld::send_chat_message(ChatMessage& message) {
     Arena arena;
     auto msg = Arena::Create<ChatMsg>(&arena);
@@ -932,6 +939,11 @@ void ClientWorld::update(float delta_time) {
     while (m_message_queue.try_pop(message)) {
         m_world_scene.handle_chat_message(message);
     }
+
+    VoiceMessage vm;
+    while (m_voice_queue.try_pop(vm)) {
+        m_audio.receive_voice(vm.data, vm.pos);
+    }
 }
 
 bool ClientWorld::handle_event(const Event& e) {

src/gameplay/network_client.cpp

@@ -141,6 +141,12 @@ asio::awaitable<void> NetworkClient::read_loop() {
                     m_world.receive_chat_message(*msg);
                 }
             } break;
+            case std::to_underlying(PacketEnum::VOICE_MSG): {
+                auto* msg = Arena::Create<VoiceMsg>(&arena);
+                if (decode_packet(*msg, body_data, header)) {
+                    m_world.receive_voice_message(*msg);
+                }
+            }
             }
         }
     } catch (const asio::system_error& e) {
@@ -228,5 +234,5 @@ void NetworkClient::set_error(std::string_view error) {
     m_error_string = error;
     m_connect_error = true;
 }
-
+ClientWorld& NetworkClient::world() { return m_world; }
 } // namespace Cubed

src/gameplay/server_world.cpp

@@ -763,6 +763,44 @@ void ServerWorld::handle_chat_message(ChatMsg& msg) {
     });
 }
 
+void ServerWorld::handle_voice_message(VoiceMsg& msg) {
+    Logger::info("Get voice Message");
+    auto pool = m_net_thread_pool.load();
+    std::string uuid = msg.uuid();
+    std::string data = msg.opus_data();
+    auto pos = msg.pos();
+    glm::vec3 p{pos.x(), pos.y(), pos.z()};
+    pool->enqueue([this, uuid = std::move(uuid), data = std::move(data), p]() {
+        std::vector<std::shared_ptr<Session>> session;
+
+        {
+            std::shared_lock lock(m_player_mutex);
+            for (auto& [key, player] : m_players) {
+                if (key == uuid) {
+                    continue;
+                }
+                session.emplace_back(player.get_session());
+            }
+        }
+
+        Arena arena;
+        auto msg = Arena::Create<VoiceMsg>(&arena);
+        msg->set_uuid(uuid);
+
+        msg->set_opus_data(data);
+
+        auto pos = msg->mutable_pos();
+        pos->set_x(p.x);
+        pos->set_y(p.y);
+        pos->set_z(p.z);
+
+        for (auto& s : session) {
+            s->send(make_packet(*msg), 5);
+        }
+        Logger::info("Send voice message sum {}", session.size());
+    });
+}
+
 void ServerWorld::handle_block_change(const BlockChangeReq& req) {
     float x = std::floor(req.pos().x());
     float y = std::floor(req.pos().y());

src/gameplay/session.cpp

@@ -105,6 +105,12 @@ asio::awaitable<void> Session::read_loop() {
                     m_server_world.handle_chat_message(*msg);
                 }
             }
+            if (cmd_id == std::to_underlying(PacketEnum::VOICE_MSG)) {
+                auto* msg = Arena::Create<VoiceMsg>(&arena);
+                if (decode_packet(*msg, body_data, header)) {
+                    m_server_world.handle_voice_message(*msg);
+                }
+            }
         }
     } catch (const asio::system_error& e) {
         auto ec = e.code();

src/proto/chat/chat.proto

@@ -1,6 +1,12 @@
 syntax = "proto3";
-
+import "common/vector3.proto";
 message ChatMsg {
     string name = 1;
     string msg = 2;
+}
+
+message VoiceMsg {
+    string uuid = 1;
+    Vec3 pos = 2;
+    bytes opus_data = 3; 
 }

src/scene/world_scene.cpp

@@ -141,7 +141,7 @@ void WorldScene::on_enter() {
         m_dev_panel.init();
         m_pasue_menu.init();
         m_hud_ui.init();
-
+        m_scene_manager.app().audio().set_client(m_client);
         m_scene_manager.app().window().set_game_running(true);
     } catch (const std::exception& e) {
         m_error_ui.set_error(e.what());
@@ -290,6 +290,18 @@ bool WorldScene::handle_key_event(const KeyEvent& e) {
         }
     }
 
+    if (e.key == Key::V) {
+        auto& recording = m_client_world.get_audio().audio_recording();
+        if (e.action == KeyAction::PRESS) {
+            recording.start();
+            return true;
+        }
+        if (e.action == KeyAction::RELEASE) {
+            recording.stop();
+            return true;
+        }
+    }
+
     if (m_paused) {
         if (m_pasue_menu.handle_event(e)) {
             return true;

src/ui/credits_ui.cpp

@@ -60,6 +60,7 @@ void CreditsUI::init() {
         add_text("protobuf");
         add_text("zstd");
         add_text("OpenAl Soft");
+        add_text("Opus");
         add_text("dr_libs");
         add_text("nlohmann/json");
         add_text("HarfBuzz");

vcpkg.json

@@ -5,6 +5,7 @@
     "openal-soft",
     "harfbuzz",
     "freetype",
+    "opus",
     {
       "name": "sdl3",
       "default-features": false