#pragma once #include "StarJson.hpp" #include "StarBiMap.hpp" #include "StarException.hpp" #include "StarGameTypes.hpp" #include "StarMaybe.hpp" #include "StarThread.hpp" #include "StarDataStreamDevices.hpp" #include "StarApplicationController.hpp" #include struct OpusDecoder; typedef std::unique_ptr OpusDecoderPtr; struct OpusEncoder; typedef std::unique_ptr OpusEncoderPtr; namespace Star { String const VoiceBroadcastPrefix = "Voice\0"s; STAR_EXCEPTION(VoiceException, StarException); enum class VoiceInputMode : uint8_t { VoiceActivity, PushToTalk }; extern EnumMap const VoiceInputModeNames; enum class VoiceChannelMode: uint8_t { Mono = 1, Stereo = 2 }; extern EnumMap const VoiceChannelModeNames; STAR_CLASS(Voice); STAR_CLASS(VoiceAudioStream); STAR_CLASS(ApplicationController); struct VoiceAudioChunk { std::unique_ptr data; size_t remaining; size_t offset = 0; VoiceAudioChunk(int16_t* ptr, size_t size) { data.reset(ptr); remaining = size; offset = 0; } inline size_t takeSamples(std::vector& out, size_t count) { size_t toRead = min(count, remaining); int16_t* start = data.get() + offset; out.insert(out.end(), start, start + toRead); offset += toRead; remaining -= toRead; return toRead; } //this one's unsafe inline int16_t takeSample() { --remaining; return *(data.get() + offset++); } inline bool exhausted() { return remaining == 0; } }; class Voice { public: // Individual speakers are represented by their connection ID. typedef ConnectionId SpeakerId; class Speaker { public: SpeakerId speakerId = 0; EntityId entityId = 0; Vec2F position = Vec2F(); String name = "Unnamed"; OpusDecoderPtr decoderMono; OpusDecoderPtr decoderStereo; VoiceAudioStreamPtr audioStream; Mutex mutex; int64_t lastReceiveTime = 0; int64_t lastPlayTime = 0; float smoothDb = -96.0f; Array dbHistory = Array::filled(0); atomic muted = false; atomic playing = 0; atomic decibelLevel = -96.0f; atomic volume = 1.0f; atomic> channelVolumes = Array::filled(1); unsigned int minimumPlaySamples = 4096; Speaker(SpeakerId speakerId); Json toJson() const; }; typedef std::shared_ptr SpeakerPtr; // Get pointer to the singleton Voice instance, if it exists. Otherwise, // returns nullptr. static Voice* singletonPtr(); // Gets reference to Voice singleton, throws VoiceException if root // is not initialized. static Voice& singleton(); Voice(ApplicationControllerPtr appController); ~Voice(); Voice(Voice const&) = delete; Voice& operator=(Voice const&) = delete; void init(); void loadJson(Json const& config, bool skipSave = false); Json saveJson() const; void save() const; void scheduleSave(); // Sets the local speaker ID and returns the local speaker. Must be called upon loading into a world. SpeakerPtr setLocalSpeaker(SpeakerId speakerId); SpeakerPtr localSpeaker(); SpeakerPtr speaker(SpeakerId speakerId); HashMap& speakers(); List sortedSpeakers(bool onlyPlaying); void clearSpeakers(); // Called when receiving input audio data from SDL, on its own thread. void readAudioData(uint8_t* stream, int len); // Called to mix voice audio with the game. void mix(int16_t* buffer, size_t frames, unsigned channels); typedef function PositionalAttenuationFunction; void update(float dt, PositionalAttenuationFunction positionalAttenuationFunction = {}); void setDeviceName(Maybe device); StringList availableDevices(); int send(DataStreamBuffer& out, size_t budget = 0); bool receive(SpeakerPtr speaker, std::string_view view); // Must be called every frame with input state, expires after 1s. void setInput(bool input = true); inline int encoderChannels() const { return (int)m_channelMode; } static OpusDecoder* createDecoder(int channels); static OpusEncoder* createEncoder(int channels); private: static Voice* s_singleton; void resetEncoder(); void resetDevice(); void openDevice(); void closeDevice(); inline bool shouldEnableInput() const { return m_enabled && m_inputEnabled; } bool playSpeaker(SpeakerPtr const& speaker, int channels); void thread(); SpeakerId m_speakerId = 0; SpeakerPtr m_clientSpeaker; HashMap m_speakers; Mutex m_activeSpeakersMutex; HashSet m_activeSpeakers; OpusEncoderPtr m_encoder; float m_outputVolume = 1.0f; float m_inputVolume = 1.0f; float m_outputAmplitude = 1.0f; float m_inputAmplitude = 1.0f; float m_threshold = -50.0f; int64_t m_lastSentTime = 0; int64_t m_lastInputTime = 0; int64_t m_lastThresholdTime = 0; int64_t m_nextSaveTime = 0; bool m_enabled = true; bool m_inputEnabled = false; bool m_loopback = false; int m_deviceChannels = 1; bool m_deviceOpen = false; Maybe m_deviceName; VoiceInputMode m_inputMode; VoiceChannelMode m_channelMode; ThreadFunction m_thread; Mutex m_threadMutex; ConditionVariable m_threadCond; atomic m_stopThread; std::vector m_decodeBuffer; std::vector m_resampleBuffer; ApplicationControllerPtr m_applicationController; struct EncodedChunk { std::unique_ptr data; size_t size; EncodedChunk(unsigned char* _data, size_t len) { data.reset(_data); size = len; } }; Mutex m_encodeMutex; std::vector m_encodedChunks; size_t m_encodedChunksLength = 0; Mutex m_captureMutex; std::queue m_capturedChunks; size_t m_capturedChunksFrames = 0; }; }