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.

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) {