diff --git "a/Docs/04-UnrealEngine/6.\345\274\200\346\224\276\344\270\226\347\225\214\345\210\266\344\275\234.md" "b/Docs/04-UnrealEngine/6.\345\274\200\346\224\276\344\270\226\347\225\214\345\210\266\344\275\234.md" index f35eee2d..40993cfa 100644 --- "a/Docs/04-UnrealEngine/6.\345\274\200\346\224\276\344\270\226\347\225\214\345\210\266\344\275\234.md" +++ "b/Docs/04-UnrealEngine/6.\345\274\200\346\224\276\344\270\226\347\225\214\345\210\266\344\275\234.md" @@ -469,7 +469,7 @@ Properties=(Path="/Script/Engine.SceneComponent:RelativeRotation",bIsPublic=Fals ![image-20231211100646090](Resources/image-20231211100646090.png) -其中 `bIsPublic` 决定了该属性被保存的值是否可以被手动修改，而不仅仅是根据流送自动序列化，其中操作的接口主要位于**ULevelStreamingPersistenceManager** 中： +其中 `bIsPublic` 决定了该属性被保存的值是否可以被手动修改，而不仅仅是根据流送自动序列化，其中操作的接口主要位于 **ULevelStreamingPersistenceManager** 中： ``` c++ class ULevelStreamingPersistenceManager : public UWorldSubsystem diff --git "a/Docs/04-UnrealEngine/7.\351\237\263\351\242\221\345\274\200\345\217\221.md" "b/Docs/04-UnrealEngine/7.\351\237\263\351\242\221\345\274\200\345\217\221.md" index 3aab9013..c00765eb 100644 --- "a/Docs/04-UnrealEngine/7.\351\237\263\351\242\221\345\274\200\345\217\221.md" +++ "b/Docs/04-UnrealEngine/7.\351\237\263\351\242\221\345\274\200\345\217\221.md" @@ -1,14 +1,14 @@ # 音频开发 -长久以来，渲染一直是图形引擎的热点，而音频似乎很难吸引开发者的注意力，但实际上，对于游戏而言，音频除了可以增加沉浸式的交互体验，还能增加不少趣味性，就像这样： +长久以来，渲染一直是图形引擎的热点，而音频似乎很难吸引开发者的注意力，但实际上，音频除了可以增加沉浸式的交互体验，还能进一步增加画面的表现力： + +https://www.bilibili.com/video/BV16C4y187Xy 笔者也正是因此，点燃了对程序开发的热情，为了让音频能有更好的图形表现效果，这才入了图形引擎的坑~ -本篇文章会从基础的音频知识开始讲解，并介绍常见的 **数字信号处理（Digital Signal Process）** 算法， - -以及最后 —— 拥抱 **Meta Sound**。 +本篇文章会从基础的音频知识开始讲解，并介绍UE5中常用的音频接口，以及一些常见的 **数字信号处理（Digital Signal Process）** 算法，并且最后 —— 拥抱 **Meta Sound** 。 > 笔者并非专业的音频开发人员，如有纰漏，烦请赐教~ @@ -16,23 +16,275 @@ 众所周知，声音是由震动产生的，就像是这样： -![Sound Waves](Resources/loudspeaker-waveform.gif) +![How Does Sound Travel From One Medium To Another?](Resources/OIP-C.jpeg) + +> 图片源自：[How Does Sound Travel From One Medium To Another? (scienceabc.com)](https://www.scienceabc.com/pure-sciences/movement-of-sound-waves-through-different-media.html) + +这里有两个视频很好地讲解了声音的本质： + +- [What is Sound-The Dr. Binocs Show](https://www.bilibili.com/video/BV16C4y187Xy) + +- [What is Sound-BranchEducation](https://www.bilibili.com/video/BV1pR4y1L7WD) + +我们通常会将声音视作一个连续的波形，就像是这样： + +![image-20240204102619133](Resources/image-20240204102619133.png) + +计算机的存储精度是有限的（0到1之间有无限个小数），如果我们想要把声音数据存储到计算机中，将会面临两个维度的精度处理问题： + +- 时间 +- 振幅 + +这会使用到 [脉冲编码调制技术（Pulse Code Modulation，简称PCM）](https://baike.baidu.com/item/脉冲编码调制/5143788?fromModule=lemma_inlink) 来将声音的 **连续信号（Continuous Signal）** 转换为电子的 **离散信号（Discrete Signal）** + +首先，我们会通过 **采样（Sampling）** 处理时间维度上的精度分割： + +![image-20240204103616957](Resources/image-20240204103616957.png) + +时间维度的采样精度参数是 **采样率（Sample Rate）** ，它代表单位时间（1s）内对声音信号的采样次数 ，常见的采样率有`44100 Hz`，`48000Hz`... + +得到采样的数据后，会进一步通过 **量化（Quantize）** 处理振幅维度上的精度分割： + +![image-20240204104205505](Resources/image-20240204104205505.png) + +振幅维度的采样精度是 **位深（Bits Per Sample）** ，它代表一个音频样本的bit存储精度，常见的位深有`8 bit`，`16 bit`，`24 bit`，`32 bit`，以 `8 bit`为例，它等价于`1字节`，能表示的数有`256(2^8)`个，算上符号位，它的数值表示范围是`[-128,127]`，可以视为它具有128个振幅梯度。 + +关于采样和量化，这里有一篇优秀的文章： + +- [音视频开发基础入门｜声音的采集与量化、音频数字信号质量、音频码率 - 知乎 (zhihu.com)](https://zhuanlan.zhihu.com/p/577850804) + +最终我们得到的数据一般称之为 **波形表/图（Wave Table）** ，如果不考虑位深的存储精度，使用浮点表示振幅，在C++中，我们可以表示为这样的存储结构： + +``` c++ +std::vector WaveTable = { + sample0,sample1,sample2,sample3,sample4... +} +``` + +打开电脑的声音设置，我们可以发现每个音频设备（输入或输出），都具有相关的配置： + +![image-20240204105850144](Resources/image-20240204105850144.png) + +在现实世界中， **音源（Sound Source）** 会向四周迸发出大量的运动粒子，一部分粒子会直接进入到人耳，还有一部分粒子会与周边物体发生一次或多次碰撞再进入到人耳中，在这种复杂的周边环境下，通常会导致人的两只耳朵会接受到不同的音频数据，这也为音频营造出了所谓的 **空间感** 。 + +因此，为了在音频存储到计算机时保留这种空间感，我们会存储多份音频数据，这个份数，我们通常称为 **通道数（Number Of Channels）** ，只有一个通道的音频我们称为 **单声道（Mono）** ，具有`2个通道`的音频我们称为 **立体声（Stereo）** 。 + +多个通道的音频样本是交叉组合在一起的，每一个组合，我们称之为 **音频块（Block）** + +对于立体声，它的存储结构如下： + +``` c++ +std::vector WaveTable = { + channel0_sample0, channel1_sample0, // Block 0 + channel0_sample1, channel1_sample1, // Block 1 + channel0_sample2, channel1_sample2, // Block 2 + channel0_sample3, channel1_sample3, // Block 3 + channel0_sample4, channel1_sample4, // Block 4 + ... +} +``` + +四声道同理： + +``` c++ +std::vector WaveTable = { + channel0_sample0, channel1_sample0, channel2_sample0, channel3_sample0, // Block 0 + channel0_sample1, channel1_sample1, channel2_sample1, channel3_sample1, // Block 1 + channel0_sample2, channel1_sample2, channel2_sample2, channel3_sample2, // Block 2 + channel0_sample3, channel1_sample3, channel2_sample3, channel3_sample3, // Block 3 + channel0_sample4, channel1_sample4, channel2_sample4, channel3_sample4, // Block 4 + ... +} +``` + +通常情况下， Wave Table 是不可以直接被音频设备播放的，因为它是一堆波形数据，而不包括音频的格式信息，即采样率，位深，通道数... + +想要它能被播放，我们还需要进一步将其编码为音频文件。 + +`*.WAV`是最常见的声音文件格式之一，它是Microsoft专为Windows开发的一种标准数字音频文件，它不同于`*.mp3`，`*.flac`，`*.aac` 等音频格式，它是一种无压缩的音频，仅仅是在WaveTable的数据之前，存储了音频的格式： + +![The structure of .wav file format. | Download Scientific Diagram](Resources/The-structure-of-wav-file-format.png) + +这里有一份完整的文档： + +- [WAV文件格式解析及处理 - 简书 (jianshu.com)](https://www.jianshu.com/p/63d7aa88582b) + +请仔细阅读上文，理解WAV格式中各个数据的意义。 + +之后，我们可以在C++中使用如下的内存对齐来描述WAV的文件头结构： + +``` c++ +struct WavFmtChunk { + uint8_t chunkID[4] = { 'f','m','t',' ' }; + int32_t chunkSize; + int16_t audioFormat; + int16_t numChannels; + int32_t sampleRate; + int32_t byteRate; + int16_t blockAlign; + int16_t bitsPerSample; +}; + +struct WavDataChunkHeader { + uint8_t chunkID[4] = { 'd','a','t','a' }; + int32_t chunkSize; +}; + +struct WavRiffChunk { + uint8_t chunkID[4] = { 'R','I','F','F' }; + int32_t chunkSize; + uint8_t format[4] = { 'W','A','V','E' }; + WavFmtChunk fmtChunk; + WavDataChunkHeader dataChunkHeader; +}; + +typedef WavRiffChunk WavHeader; +``` + +在C++中存储一段数据，我们通常会使用`std::vector` ，如果想要将一段波形数据写入到Wav文件中，就可以使用这样的函数： + +``` c++ +void writeWavFile(std::string filePath, int numChannels, int sampleRate, int bitsPerSample, std::vector waveTable) { + WavHeader header; + header.fmtChunk.audioFormat = 0x0001; // PCM格式对应的数值为0x0001 + header.fmtChunk.chunkSize = 16; // PCM格式的音频格式占16字节，如果是其他格式，可能会>16，将导致总的文件头尺寸>44 + header.fmtChunk.numChannels = numChannels; + header.fmtChunk.bitsPerSample = bitsPerSample; + header.fmtChunk.sampleRate = sampleRate; + header.fmtChunk.blockAlign = header.fmtChunk.numChannels * header.fmtChunk.bitsPerSample / 8; + header.fmtChunk.byteRate = header.fmtChunk.sampleRate * header.fmtChunk.blockAlign; + header.dataChunkHeader.chunkSize = waveTable.size(); + header.chunkSize = sizeof(header.format) + sizeof(header.fmtChunk) + sizeof(header.dataChunkHeader) + waveTable.size(); + std::ofstream out(filePath, std::ios::out | std::ios::binary); //务必使用std::ios::binary，否则数据错乱 + assert(out.is_open() && "Can`t Write File"); + out.write(reinterpret_cast(&header), sizeof(WavHeader)); + out.write(reinterpret_cast(waveTable.data()), waveTable.size()); + out.close(); +} + +``` + +读取也非常简单： + +``` c++ +std::vector readWavFile(std::string filePath) { + WavHeader header; + std::ifstream in(filePath, std::ios::in | std::ios::binary); //务必使用std::ios::binary，否则数据错乱 + assert(in.is_open() && "Can`t Read File"); + in.read(reinterpret_cast(&header), sizeof(header)); + int waveTableSize = header.dataChunkHeader.chunkSize; //获取到波形数据的大小 + std::vector waveTable(waveTableSize); + in.read(reinterpret_cast(waveTable.data()), waveTableSize); + return waveTable; +} +``` + +接下来我们来生成一段音频波形， **正弦波（Sine Wave）** 是基本的波形之一，它的函数式如下： +$$ +y(Time) = Amplitude * sin(2 * PI * Frequency * Time + PhaseOffset) + Bias +$$ +根据函数式，我们可以编写如下的函数： + +```C++ +const double M_PI = 3.1415926535; +std::vector generateSineWave(int timeLengthSec, // 正弦波的时长，单位为秒 + int frequency, // 正弦波的频率 + float amplitude, // 正弦波的振幅 + int numChannels, // 音频的通道数 + int sampleRate, // 音频的采样率 + int bitsPerSample // 音频的位深 + ) { + int numBlocks = timeLengthSec * sampleRate; // 音频块的数量 + amplitude = std::clamp(amplitude, 0.01f, 1.0f); // 限制振幅的大小, std::clamp 需要C++17 + std::vector waveTable; + waveTable.resize(numBlocks * numChannels * bitsPerSample / 8); // 算出音频数据所需的内存大小 + int step = bitsPerSample / 8; + for (int i = 0; i < numBlocks; i++) { + float time = i / (float)numBlocks * timeLengthSec; // 时间点 + float sampleValue = amplitude * std::sin(2 * M_PI * frequency * time); //算出对应时间点的采样值 + // 我们得想个办法将sampleValue写入到waveTable里面 + } + return waveTable; +} +``` + +上面的代码里面我们面临一个难题： -这里有一个通俗的视频讲解了声音的本质： +- 通过正弦波的函数，我们会得到一个类型为`float`的采样值，但位深可能是`8 bit`，`16 bit`，`24 bit`，`32 bit`，我们该怎么把float转换为对应位深的数据呢？ - +常见的做法借助已有的数据类型来进行对应数据位的赋值，就像是这样： +``` c++ +const int PCMS16MaxAmplitude = 32767; // 16 bit的数值范围是[-32768,32767]，这里取最小绝对值（32767）避免越界 +void write16BitSampleValue(uint8_t* dataPtr, float sampleValue) { + (*reinterpret_cast(dataPtr)) = sampleValue * PCMS16MaxAmplitude; +} +``` -https://zhuanlan.zhihu.com/p/550072985 +也可以利用位运算来进行填充： -[有哪些国内的游戏使用了音频引擎来做音频部分的开发？比如FMOD和wwise? - 知乎 (zhihu.com)](https://www.zhihu.com/question/29992299) +``` c++ +const int PCMS24MaxAmplitude = 8388607; +void write24BitSampleValue(uint8_t* dataPtr, float sampleValue) { + int value = sampleValue * PCMS24MaxAmplitude; + dataPtr[0] = value & 0xFF; + dataPtr[1] = (value >> 8) & 0xFF; + dataPtr[2] = (value >> 16) & 0xFF; +} +``` -[Sounds Good: Unreal Engine Audio | Unreal Fest 2023](https://www.bilibili.com/video/BV1EC4y1J7Bs) +综上，我们可以整理出一个这样的转换函数表： + +``` c++ +const int PCMS8MaxAmplitude = 127; +const int PCMS16MaxAmplitude = 32767; +const int PCMS24MaxAmplitude = 8388607; +const int PCMS32MaxAmplitude = 2147483647; + +typedef std::function SampleValueWriteFunction; +static std::map SampleValueWriteFunctions = { + {8,[](uint8_t* dataPtr, float sampleValue) { + // 笔者不清楚为什么这里需要加上一个偏差波形才是正确的=.= + (*reinterpret_cast(dataPtr)) = sampleValue * PCMS8MaxAmplitude + PCMS8MaxAmplitude; + }}, + {16,[](uint8_t* dataPtr, float sampleValue) { + (*reinterpret_cast(dataPtr)) = sampleValue * PCMS16MaxAmplitude; + }}, + {24,[](uint8_t* dataPtr, float sampleValue) { + int value = sampleValue * PCMS24MaxAmplitude; + dataPtr[0] = value & 0xFF; + dataPtr[1] = (value >> 8) & 0xFF; + dataPtr[2] = (value >> 16) & 0xFF; + }}, + {32,[](uint8_t* dataPtr, float sampleValue) { + (*reinterpret_cast(dataPtr)) = sampleValue * PCMS32MaxAmplitude; + }} +}; -[WAV文件格式解析及处理 - 简书 (jianshu.com)](https://www.jianshu.com/p/63d7aa88582b) +typedef std::function SampleValueReadFunction; +static std::map SampleValueReadFunctions = { + {8,[](uint8_t* dataPtr) { + return ((*reinterpret_cast(dataPtr)) - PCMS8MaxAmplitude) / float(PCMS8MaxAmplitude); + }}, + {16,[](uint8_t* dataPtr) { + return (*reinterpret_cast(dataPtr)) / float(PCMS16MaxAmplitude); + }}, + {24,[](uint8_t* dataPtr) { + int value = dataPtr[2]; + value <<= 8; + value |= dataPtr[1]; + value <<= 8; + value |= dataPtr[0]; + return value / float(PCMS24MaxAmplitude); + }}, + {32,[](uint8_t* dataPtr) { + return (*reinterpret_cast(dataPtr)) / float(PCMS32MaxAmplitude); + }} +}; +``` -[Musiscope (sciencemusic.org)](https://oscilloscope.sciencemusic.org/) +综上，加上测试代码，完整的代码如下： ``` c++ #include @@ -45,8 +297,8 @@ https://zhuanlan.zhihu.com/p/550072985 #include #include -struct WaveFmtChunk { - unsigned char chunkID[4] = { 'f','m','t',' ' }; +struct WavFmtChunk { + uint8_t chunkID[4] = { 'f','m','t',' ' }; int32_t chunkSize; int16_t audioFormat; int16_t numChannels; @@ -56,54 +308,54 @@ struct WaveFmtChunk { int16_t bitsPerSample; }; -struct WaveDataChunkHeader { - unsigned char chunkID[4] = { 'd','a','t','a' }; +struct WavDataChunkHeader { + uint8_t chunkID[4] = { 'd','a','t','a' }; int32_t chunkSize; }; -struct WaveRiffChunk { - unsigned char chunkID[4] = { 'R','I','F','F' }; +struct WavRiffChunk { + uint8_t chunkID[4] = { 'R','I','F','F' }; int32_t chunkSize; - unsigned char format[4] = { 'W','A','V','E' }; - WaveFmtChunk fmtChunk; - WaveDataChunkHeader dataChunkHeader; + uint8_t format[4] = { 'W','A','V','E' }; + WavFmtChunk fmtChunk; + WavDataChunkHeader dataChunkHeader; }; -typedef WaveRiffChunk WaveHeader; +typedef WavRiffChunk WavHeader; -const int PCMS8MaxAmplitude = 127; +const int PCMS8MaxAmplitude = 127; const int PCMS16MaxAmplitude = 32767; const int PCMS24MaxAmplitude = 8388607; const int PCMS32MaxAmplitude = 2147483647; -typedef std::function SampleValueWriteFunction; -static std::map SampleValueWriteFunctions = { - {8,[](unsigned char* dataPtr, float sampleValue) { - (*reinterpret_cast(dataPtr)) = sampleValue * PCMS8MaxAmplitude + PCMS8MaxAmplitude; +typedef std::function SampleValueWriteFunction; +static std::map SampleValueWriteFunctions = { + {8,[](uint8_t* dataPtr, float sampleValue) { + (*reinterpret_cast(dataPtr)) = sampleValue * PCMS8MaxAmplitude + PCMS8MaxAmplitude; // 笔者不清楚为什么这里需要加上一个偏差波形才是正确的=.= }}, - {16,[](unsigned char* dataPtr, float sampleValue) { + {16,[](uint8_t* dataPtr, float sampleValue) { (*reinterpret_cast(dataPtr)) = sampleValue * PCMS16MaxAmplitude; }}, - {24,[](unsigned char* dataPtr, float sampleValue) { + {24,[](uint8_t* dataPtr, float sampleValue) { int value = sampleValue * PCMS24MaxAmplitude; dataPtr[0] = value & 0xFF; dataPtr[1] = (value >> 8) & 0xFF; dataPtr[2] = (value >> 16) & 0xFF; }}, - {32,[](unsigned char* dataPtr, float sampleValue) { + {32,[](uint8_t* dataPtr, float sampleValue) { (*reinterpret_cast(dataPtr)) = sampleValue * PCMS32MaxAmplitude; }} }; -typedef std::function SampleValueReadFunction; -static std::map SampleValueReadFunctions = { - {8,[](unsigned char* dataPtr) { +typedef std::function SampleValueReadFunction; +static std::map SampleValueReadFunctions = { + {8,[](uint8_t* dataPtr) { return ((*reinterpret_cast(dataPtr)) - PCMS8MaxAmplitude) / float(PCMS8MaxAmplitude); }}, - {16,[](unsigned char* dataPtr) { + {16,[](uint8_t* dataPtr) { return (*reinterpret_cast(dataPtr)) / float(PCMS16MaxAmplitude); }}, - {24,[](unsigned char* dataPtr) { + {24,[](uint8_t* dataPtr) { int value = dataPtr[2]; value <<= 8; value |= dataPtr[1]; @@ -111,80 +363,757 @@ static std::map SampleValueReadFunctions value |= dataPtr[0]; return value / float(PCMS24MaxAmplitude); }}, - {32,[](unsigned char* dataPtr) { + {32,[](uint8_t* dataPtr) { return (*reinterpret_cast(dataPtr)) / float(PCMS32MaxAmplitude); }} }; -const double MATH_PI = 3.1415926535; +const double M_PI = 3.1415926535; -std::vector generateSineWave(int timeLengthSec, int frequency, float amplitude, int numChannels, int sampleRate, int bitsPerSample) { - int numBlocks = timeLengthSec * sampleRate; - amplitude = std::clamp(amplitude, 0.01f, 1.0f); - std::vector audioData; - audioData.resize(numBlocks * numChannels * bitsPerSample / 8); +std::vector generateSineWave(int timeLengthSec, int frequency, float amplitude, int numChannels, int sampleRate, int bitsPerSample) { + int numBlocks = timeLengthSec * sampleRate; // 音频块的数量 + amplitude = std::clamp(amplitude, 0.01f, 1.0f); // 限制振幅的大小, std::clamp 需要C++17 + std::vector waveTable; + waveTable.resize(numBlocks * numChannels * bitsPerSample / 8); int step = bitsPerSample / 8; - unsigned char* audioDataPtr = audioData.data(); + uint8_t* waveTablePtr = waveTable.data(); SampleValueWriteFunction writer = SampleValueWriteFunctions.at(bitsPerSample); - std::string before; for (int i = 0; i < numBlocks; i++) { - double time = i / (double)numBlocks * timeLengthSec; - float sampleValue = amplitude * std::sin(2 * MATH_PI * frequency * time); - for (int j = 0; j < numChannels; j++) { - writer(audioDataPtr, sampleValue); - audioDataPtr += step; + float time = i / (float)numBlocks * timeLengthSec; + float sampleValue = amplitude * std::sin(2 * M_PI * frequency * time); + for (int channelIndex = 0; channelIndex < numChannels; channelIndex++) { + writer(waveTablePtr, sampleValue); + waveTablePtr += step; } } - return audioData; -} - -void writeWaveFile(std::string filePath, int numChannels, int sampleRate, int bitsPerSample, std::vector audioData) { - WaveHeader header; - header.fmtChunk.audioFormat = 0x0001; // PCM格式对应的数值为0x0001 - header.fmtChunk.chunkSize = 16; // PCM格式的音频格式占16字节，如果是其他格式，可能会>16,即总的文件头大小>44 - header.fmtChunk.numChannels = numChannels; - header.fmtChunk.bitsPerSample = bitsPerSample; - header.fmtChunk.sampleRate = sampleRate; - header.fmtChunk.blockAlign = header.fmtChunk.numChannels * header.fmtChunk.bitsPerSample / 8; - header.fmtChunk.byteRate = header.fmtChunk.sampleRate * header.fmtChunk.blockAlign; - header.dataChunkHeader.chunkSize = audioData.size(); - header.chunkSize = sizeof(header.format) + sizeof(header.fmtChunk) + sizeof(header.dataChunkHeader) + audioData.size(); + return waveTable; +} + +void writeWavFile(std::string filePath, int numChannels, int sampleRate, int bitsPerSample, std::vector waveTable) { + WavHeader header; + header.fmtChunk.audioFormat = 0x0001; // PCM格式对应的数值为0x0001 + header.fmtChunk.chunkSize = 16; // PCM格式的音频格式占16字节，如果是其他格式，可能会>16，将导致总的文件头尺寸>44 + header.fmtChunk.numChannels = numChannels; + header.fmtChunk.bitsPerSample = bitsPerSample; + header.fmtChunk.sampleRate = sampleRate; + header.fmtChunk.blockAlign = header.fmtChunk.numChannels * header.fmtChunk.bitsPerSample / 8; + header.fmtChunk.byteRate = header.fmtChunk.sampleRate * header.fmtChunk.blockAlign; + header.dataChunkHeader.chunkSize = waveTable.size(); + header.chunkSize = sizeof(header.format) + sizeof(header.fmtChunk) + sizeof(header.dataChunkHeader) + waveTable.size(); std::ofstream out(filePath, std::ios::out | std::ios::binary); //务必使用std::ios::binary，否则数据错乱 - out.write(reinterpret_cast(&header), sizeof(WaveHeader)); - out.write(reinterpret_cast(audioData.data()), audioData.size()); + assert(out.is_open() && "Can`t Write File"); + out.write(reinterpret_cast(&header), sizeof(WavHeader)); + out.write(reinterpret_cast(waveTable.data()), waveTable.size()); out.close(); } -std::vector readWaveFile(std::string filePath) { - std::ifstream in(filePath , std::ios::in | std::ios::binary); //务必使用std::ios::binary，否则数据错乱 - WaveHeader header; +std::vector readWavFile(std::string filePath) { + WavHeader header; + std::ifstream in(filePath, std::ios::in | std::ios::binary); //务必使用std::ios::binary，否则数据错乱 + assert(in.is_open() && "Can`t Read File"); in.read(reinterpret_cast(&header), sizeof(header)); - std::vector audioData(header.dataChunkHeader.chunkSize); - in.read(reinterpret_cast(audioData.data()), header.dataChunkHeader.chunkSize); - return audioData; + int waveTableSize = header.dataChunkHeader.chunkSize; //获取到波形数据的大小 + std::vector waveTable(waveTableSize); + in.read(reinterpret_cast(waveTable.data()), waveTableSize); + return waveTable; } int main() { - assert(sizeof(WaveHeader) == 44 && "Wave Header Size Error"); + assert(sizeof(WavHeader) == 44 && "WAV Header Size Error"); - int numChannels = 1; //单通道 + int numChannels = 2; //单通道 int sampleRate = 48000; //采样率为48000HZ int bitsPerSample = 16; //每个音频样本占16Bit，即2字节 int auidoLengthSec = 2; //生成2s的正弦波 int sineFrequency = 440; //正弦波的频率 float sineAmplitude = 0.8f; //正弦波的振幅 - std::vector audioData = generateSineWave(auidoLengthSec, sineFrequency, sineAmplitude, numChannels, sampleRate, bitsPerSample); + std::vector waveTable = generateSineWave(auidoLengthSec, sineFrequency, sineAmplitude, numChannels, sampleRate, bitsPerSample); + + writeWavFile("./sineWave.wav", numChannels, sampleRate, bitsPerSample, waveTable); + std::vector readData = readWavFile("./sineWave.wav"); + assert(readData == waveTable && "Audio Data Error"); - writeWaveFile("./test.raw", numChannels, sampleRate, bitsPerSample, audioData); + return 0; +} +``` - std::vector readData = readWaveFile("./test.wav"); +执行该程序会在输出目录生成一个`sineWave.wav`音频文件，尝试播放它~ - assert(readData == audioData && "Audio Data Error"); +![image-20240204141548048](Resources/image-20240204141548048.png) - return 0; +你还可以修改一下参数，重新生成音频，听听不同频率下的正弦波有什么区别 + +在这个网站上，可以上传音频文件，查看它的波形： + +- [Musiscope (sciencemusic.org)](https://oscilloscope.sciencemusic.org/) + +![image-20240204142046638](Resources/image-20240204142046638.png) + +![image-20240204142115830](Resources/image-20240204142115830.png) + +至此，我们应该大致搞明白了音频在计算机上的原始形态。 + +## 音频框架 + +### 音频IO接口 + +有了音频，我们还需要清楚，怎么跟音频设备打交道，而常见的音频设备有： + +- 输入：麦克风 +- 输出：扬声器，音响，耳机 + +这些设备厂商会提供或支持相关的音频驱动，通常情况下，我们无需直接跟音频驱动打交道，而是利用操作系统提供的音频接口来处理音频的输入输出，例如： + +- **Windows：** DirectSound，ASIO，WASAPI +- **Linux： ** native ALSA，Jack， OSS +- **Mac OS：** CoreAudio ，Jack + +为了跨平台的音频处理，很多框架也会对这些接口进行统一的封装，就比如： + +- **Qt Multimedia** 中的音频IO模块：[Audio Overview | Qt Multimedia 6.6.1](https://doc.qt.io/qt-6/audiooverview.html) +- **UE** 引入的 **RtAudio** ： http://www.music.mcgill.ca/~gary/rtaudio/ + +### 傅里叶变换库 + +傅里叶变换（Fourier Transform）是信号处理的重要工具。 + +已知声音是一段随时间变换的连续信号，就像是这样： + +![image-20240205162859891](Resources/image-20240205162859891.png) + +但我们听到的声音往往不会只有一种频率，实际上，可能会是由很多频率的声波堆叠到一起，就像是这样： + +![image-20240205162920488](Resources/image-20240205162920488.png) + +而傅里叶能帮助我们，分析 **一段时间内** 已经混合了大量不同频率波形 的 **时域信号（Time Domain Signal）** ，得出该时间段内声音波形的频率分布，即 **频域信号（Frequency Domain Signal）** ，整个过程看起来就像是这样的： + +![TikZ 绘制傅里叶级数示意图 - LaTeX 工作室](Resources/9796875423a0a9928413d91f8c7420c2.png) + +这里有一个非常好的视频讲解了傅里叶变换的原理： + +- [【官方双语】形象展示傅里叶变换 |3Blue1Brown](https://www.bilibili.com/video/BV1pW411J7s8) + +一些看起来比较混乱的时序信号，在转换到频域之后，我们会更容易发现数据的特征，就比如这是一段音乐的实时分析结果： + +> 虽然听不到声音，但相信你也能从跃动的图形能看出音乐的节奏变化。 + +![sfsa](Resources/sfsa.gif) + +此外，傅里叶变换不仅仅是一个数学处理的工具，它还引导我们在频域去思考问题，如果你喜欢玩 Shader Toy 或 GLSL Sandbox，里面很多代码其实都是基于频域思考的。 + +这里有一个有趣的视频，每条线都在按固定的频率匀速转动： + +- [傅里叶变换之美 | 科学小视](https://www.bilibili.com/video/BV1So4y1T7ae) + +这里有一些比较成熟的傅里叶算法实现： + +- [fftw3](https://github.com/FFTW/fftw3)：西方最快的傅里叶变换库 +- [pffft](https://bitbucket.org/jpommier/pffft)：一个小巧的傅里叶变换库 +- [ffts](https://github.com/anthonix/ffts)：南方最快的傅里叶变换库（狗头） + +如果你也像笔者一样喜欢音频可视化，这里推荐两个开源库： + +- [adamstark/Gist: A C++ Library for Audio Analysis (github.com)](https://github.com/adamstark/Gist) +- [jmerkt/rt-cqt: Real-Time Constant-Q Transform (github.com)](https://github.com/jmerkt/rt-cqt) + +### DSP算法库 + +DSP的全称是 **数字信号处理（Digital Signal Processor）** + +在 `认识音频` 的小节中，我们生成了一个正弦波，并写入到wav文件，通常情况下，这些功能会囊括在DSP库中，此外，它往往还会提供一些其他功能，例如： + +- 滤波器 +- 重采样 +- 音频合成 +- 频谱分析 +- ... + +常见的开源DSP库有： + +- [KFR | Fast, modern C++ DSP framework (kfrlib.com)](https://www.kfrlib.com/) +- [Maximilian | C++ Audio and Music DSP Library](https://github.com/micknoise/Maximilian) + +### 音频开发引擎 + +对于游戏项目的音频开发人员，通常情况下，无需了解上面的一些底层接口和框架，一般游戏引擎都会提供基本的音频模块，涉及到复杂的音频机制，我们一般会使用音频引擎来进行开发。 + +常见的音频引擎有： + +- [CRIWARE ADX](https://www.criware.com/en/products/adx2.html)：ADX 几乎没有学习曲线，因为它的创作工具充分利用了 DAW 范式所提供的功能，并添加了直观的功能来创建交互式声音。在项目规模越来越大、期限越来越紧的时候，使用 ADX 提高您的工作效率！ +- [FMOD](https://fmod.com/)：FMOD 是一种端到端解决方案，用于为任何游戏添加声音和音乐。使用 FMOD Studio 构建自适应音频，并使用 FMOD 引擎在游戏中播放。 +- [Wwise | Audiokinetic](https://www.audiokinetic.com/zh/products/wwise/)：可提供一整套互动音频设计与开发工具，无论项目规模如何都能帮您轻松完成原型设计，并在此基础上将自己的音频创作构想变为现实。 + +这些音频软件都会提供独特的编辑器，并在相关的游戏引擎（UE，Unity，Cocos...）中提供插件支持。 + +关于音频引擎的选择，知乎上面有一些相关的讨论： + +- [有哪些国内的游戏使用了音频引擎来做音频部分的开发？比如FMOD和wwise? - 知乎 (zhihu.com)](https://www.zhihu.com/question/29992299) + +- [使用Unity5开发手游，使用FMOD或wwise音频插件而不是自带的声音系统有哪些好处？ - 知乎 (zhihu.com)](https://www.zhihu.com/question/61882604) + +**Meta Sound** 算是UE5提供的一个内置的音频引擎，虽然目前它的功能并不能媲美上面的几个音频引擎，但这种一站式的解决方案，无疑会让UE变得越来越强大。 + +就像 UE5 的建模工具一样，它最大的意义不在于让美术人员也能在UE里进行建模，而是开发人员可以很方便地利用UE底层提供的一系列网格处理算法，根据自己项目的需求搭建出一条从DCC模型到高品质引擎资产的自动化管线。 + +## UE 音频 + + 在 Unreal Fest 2023 ，大佬介绍了 UE 的音频架构： + +- [Sounds Good: Unreal Engine Audio | Unreal Fest 2023](https://www.bilibili.com/video/BV1EC4y1J7Bs) + +![image-20240204155736479](Resources/image-20240204155736479.png) + +我们能从如下站点获取到一些信息，但可惜的是，虚幻中关于音频的文档相对而言比较少，并且混杂着大量已过时的文档： + +- [在虚幻引擎中使用音频 |虚幻引擎 5.3 文档 (unrealengine.com)](https://docs.unrealengine.com/5.3/en-US/working-with-audio-in-unreal-engine/) +- [Latest unreal-engine topics in Audio - Epic Developer Community Forums (unrealengine.com)](https://forums.unrealengine.com/tags/c/development-discussion/audio/42/unreal-engine) + +### 音频设备 + +音频设备分为输入设备和输出设备，它们的相关信息分别存储在UE的 **Audio::FCaptureDeviceInfo** 和 **Audio::FAudioPlatformDeviceInfo** ，我们可以通过如下接口获取到所有的设备： + +``` c++ +// 获取所有的输入设备 +Audio::FAudioCapture AudioCapture; +TArray CaptureDevices; +AudioCapture.GetCaptureDevicesAvailable(CaptureDevices); +for (auto InputDeviceInfo : CaptureDevices) { + UE_LOG(LogTemp, Warning,TEXT("[Input Device] Name: %s, Device Id: %s, Num Channels: %u, Sample Rate: %u, Supports Hardware AEC: %u, "), + *InputDeviceInfo.DeviceName, + *InputDeviceInfo.DeviceId, + InputDeviceInfo.InputChannels, + InputDeviceInfo.PreferredSampleRate, + InputDeviceInfo.bSupportsHardwareAEC); +} + +// 获取当前（默认）音频输入设备 +Audio::FCaptureDeviceInfo CurrentInputDevice; +AudioCapture.GetCaptureDeviceInfo(CurrentInputDevice); +UE_LOG(LogTemp, Warning, TEXT("[Current Input Device] %s"), + *CurrentInputDevice.DeviceName); + +// 获取所有的输出设备 +Audio::FMixerDevice* AudioMixerDevice = FAudioDeviceManager::GetAudioMixerDeviceFromWorldContext(GWorld); //使用当前GWorld作为WorldContentObject +if (AudioMixerDevice){ + if (Audio::IAudioMixerPlatformInterface* MixerPlatform = AudioMixerDevice->GetAudioMixerPlatform()){ + uint32 NumOutputDevices = 0; + MixerPlatform->GetNumOutputDevices(NumOutputDevices); + for (uint32 i = 0; i < NumOutputDevices; ++i){ + Audio::FAudioPlatformDeviceInfo OutputDeviceInfo; + MixerPlatform->GetOutputDeviceInfo(i, OutputDeviceInfo); + UE_LOG(LogTemp, Warning, TEXT("[Output Device] Name: %s, Device Id: %s, Num Channels: %u, Sample Rate: %u, "), + *OutputDeviceInfo.Name, + *OutputDeviceInfo.DeviceId, + OutputDeviceInfo.NumChannels, + OutputDeviceInfo.SampleRate); + } + + // 获取当前（默认）音频输出设备 + Audio::FAudioPlatformDeviceInfo CurrentOutputDevice = MixerPlatform->GetPlatformDeviceInfo(); + UE_LOG(LogTemp, Warning, TEXT("[Current Output Device] %s"), + *CurrentOutputDevice.Name); + } +} +``` + +![image-20240204170415081](Resources/image-20240204170415081.png) + +## 输出音频 + +UE 中的音源基类是 **USoundBase** ： + +![image-20240204191436270](Resources/image-20240204191436270.png) + +引擎中有许多派生： + +![image-20240204191454472](Resources/image-20240204191454472.png) + +我们的关注点主要在 **USoundWave** 上，它会将波形数据传输给音频设备进行播放。 + +结合我们之前编写的正弦波生成函数，我们也可以在UE里播放我们自定义的波形数据，就像是这样： + +``` c++ +void UTestBlueprintLibrary::PlayTestAudio(UObject* WorldContext) // 用于测试的蓝图函数库 +{ + int numChannels = 2; //单通道 + int sampleRate = 48000; //采样率为48000HZ + int bitsPerSample = 16; //每个音频样本占16Bit，即2字节 + int auidoLengthSec = 2; //生成2s的正弦波 + int sineFrequency = 440; //正弦波的频率 + float sineAmplitude = 0.8f; //正弦波的振幅 + + USoundWaveProcedural* Wave = NewObject(); + Wave->NumChannels = numChannels; + Wave->SampleByteSize = bitsPerSample / 8; // UE只支持int16 + Wave->SetSampleRate(sampleRate); + std::vector waveTable = generateSineWave(auidoLengthSec, + sineFrequency, + sineAmplitude, + numChannels, + sampleRate, + bitsPerSample); + Wave->QueueAudio(waveTable.data(), waveTable.size()); + UGameplayStatics::PlaySound2D(WorldContext, Wave); +} +``` + +这里用到了函数`UGameplayStatics::PlaySound2D`，我们可以看下它的内部实现： + +``` c++ +void UGameplayStatics::PlaySound2D(const UObject* WorldContextObject, + USoundBase* Sound, + float VolumeMultiplier, + float PitchMultiplier, + float StartTime, + USoundConcurrency* ConcurrencySettings, + const AActor* OwningActor, + bool bIsUISound) +{ + if (!Sound || !GEngine || !GEngine->UseSound()) + { + return; + } + + UWorld* ThisWorld = GEngine->GetWorldFromContextObject(WorldContextObject, EGetWorldErrorMode::LogAndReturnNull); + if (!ThisWorld || !ThisWorld->bAllowAudioPlayback || ThisWorld->IsNetMode(NM_DedicatedServer)) + { + return; + } + + if (FAudioDeviceHandle AudioDevice = ThisWorld->GetAudioDevice()) // 获取当前世界的音频输出设备 + { + FActiveSound NewActiveSound; // 创建一个激活的音频 + NewActiveSound.SetSound(Sound); + NewActiveSound.SetWorld(ThisWorld); + + NewActiveSound.SetPitch(PitchMultiplier); + NewActiveSound.SetVolume(VolumeMultiplier); + + NewActiveSound.RequestedStartTime = FMath::Max(0.f, StartTime); + + NewActiveSound.bIsUISound = bIsUISound; + NewActiveSound.bAllowSpatialization = false; + + if (ConcurrencySettings) + { + NewActiveSound.ConcurrencySet.Add(ConcurrencySettings); + } + + NewActiveSound.Priority = Sound->Priority; + NewActiveSound.SubtitlePriority = Sound->GetSubtitlePriority(); + + // If OwningActor isn't supplied to this function, derive an owner from the WorldContextObject + const AActor* ActiveSoundOwner = OwningActor ? OwningActor : GameplayStatics::GetActorOwnerFromWorldContextObject(WorldContextObject); + + NewActiveSound.SetOwner(ActiveSoundOwner); + + TArray Params; + UActorSoundParameterInterface::Fill(ActiveSoundOwner, Params); + + AudioDevice->AddNewActiveSound(NewActiveSound, &Params); // 将激活音频扔给输出设备处理 + } +} +``` + +我们调用`USoundWaveProcedural::QueueAudio`添加的Wave数据，会存储到内部的AudioBuffer中，后续会被音频设备调用`USoundWaveProcedural::GeneratePCMData`来接受，并播放声音。 + +``` c++ +int32 USoundWaveProcedural::GeneratePCMData(uint8* PCMData, const int32 SamplesNeeded) +{ + // ... + FMemory::Memcpy((void*)PCMData, &AudioBuffer[0], BytesToCopy); +} +``` + +使用 **USoundWaveProcedural** 我们可以播放PCM数据，但如果涉及到一些复杂的音频生成策略，我们可能需要派生 **USoundWaveProcedural** ， **UMetaSoundSource** 的实现是一个很好的参考示例，我们只需自定义一个 **ISoundGenerator** ，覆写它的音频生成函数： + +``` c++ +class ISoundGenerator +{ +public: + ENGINE_API ISoundGenerator(); + ENGINE_API virtual ~ISoundGenerator(); + + // Called when a new buffer is required. + virtual int32 OnGenerateAudio(float* OutAudio, int32 NumSamples) = 0; + + // Returns the number of samples to render per callback + virtual int32 GetDesiredNumSamplesToRenderPerCallback() const { return 1024; } + + // Optional. Called on audio generator thread right when the generator begins generating. + virtual void OnBeginGenerate() {} + + // Optional. Called on audio generator thread right when the generator ends generating. + virtual void OnEndGenerate() {} + + // Optional. Can be overridden to end the sound when generating is finished. + virtual bool IsFinished() const { return false; } + + // Retrieves the next buffer of audio from the generator, called from the audio mixer + ENGINE_API int32 GetNextBuffer(float* OutAudio, int32 NumSamples, bool bRequireNumberSamples = false); + + virtual Audio::AudioTaskQueueId GetSynchronizedRenderQueueId() const { return 0; } + +protected: + + // Protected method to execute lambda in audio render thread + // Used for conveying parameter changes or events to the generator thread. + ENGINE_API void SynthCommand(TUniqueFunction Command); + +private: + + ENGINE_API void PumpPendingMessages(); + + // The command queue used to convey commands from game thread to generator thread + TQueue> CommandQueue; + + friend class USynthComponent; +}; +``` + +再派生自己的 **USoundWaveProcedural** ，覆写： + +``` c++ +virtual ISoundGeneratorPtr USoundBase::CreateSoundGenerator(const FSoundGeneratorInitParams& InParams) { return nullptr; } +``` + +UE的音频设备在播放音频源时，如果它是程序化音频`(USoundWave.bProcedural == true)`，那么将会调用`CreateSoundGenerator`为其创建 **SoundGenerator** ： + +![image-20240204195219544](Resources/image-20240204195219544.png) + +在实际解析音频时，会调用如下接口生成音频： + +![image-20240204195409513](Resources/image-20240204195409513.png) + +详细的实现可参考 **UMetaSoundSource** 和 **FMetasoundGenerator** + +### 捕获音频 + +游戏项目中很少会有音频捕获的需求，因此这里只做简单介绍。 + +关于音频捕获，主要分为两类： + +- 捕获系统的音频 +- 捕获游戏内的音频 + +#### 系统音频捕获 + +系统音频捕获 特指 输入设备（麦克风）的采集，UE目前并不支持输出设备的捕获，如果有相应的需求，则需要使用更低级别的API来实现（如WASAPI）。 + +``` c++ +TSharedPtr AudioCapture = MakeShared(); +Audio::FOnCaptureFunction OnCapture = [](const float* AudioData, + int32 NumFrames, + int32 InNumChannels, + int32 InSampleRate, + double StreamTime, + bool bOverFlow) { + // 处理采集到的音频数据，此时是在异步线程 +}; +Audio::FAudioCaptureDeviceParams Params; // 音频采集的参数配置，不填设备则会使用默认输入设备 +if (mAudioCapture->OpenCaptureStream(Params, MoveTemp(OnCapture), 1024)) { // 开始音频捕获 + Audio::FCaptureDeviceInfo Info; + if (mAudioCapture->GetCaptureDeviceInfo(Info)){ + OnAudioFormatChanged.ExecuteIfBound(Info.InputChannels, Info.PreferredSampleRate); // 音频格式变换的通知信号 + } } ``` +#### 游戏内音频捕获 + +游戏内音频捕获使用UE提供的接口 **ISubmixBufferListener** ： + +``` c++ +/** Abstract interface for receiving audio data from a given submix. */ +class ISubmixBufferListener +{ +public: + virtual ~ISubmixBufferListener() = default; + + /** + Called when a new buffer has been rendered for a given submix + @param OwningSubmix The submix object which has rendered a new buffer + @param AudioData Ptr to the audio buffer + @param NumSamples The number of audio samples in the audio buffer + @param NumChannels The number of channels of audio in the buffer (e.g. 2 for stereo, 6 for 5.1, etc) + @param SampleRate The sample rate of the audio buffer + @param AudioClock Double audio clock value, from start of audio rendering. + */ + virtual void OnNewSubmixBuffer(const USoundSubmix* OwningSubmix, float* AudioData, int32 NumSamples, int32 NumChannels, const int32 SampleRate, double AudioClock) = 0; + + /** + * Called if the submix is evaluating disabling itself for the next buffer of audio in FMixerSubmix::IsRenderingAudio() + * if this returns true, FMixerSubmix::IsRenderingAudio() will return true. + * Otherwise the submix will evaluate playing sounds, children submixes, etc to see if it should auto-disable + * + * This is called every evaluation. + * ISubmixListeners that intermittently render audio should only return true when they have work to do. + */ + virtual bool IsRenderingAudio() const + { + return false; + } +}; +``` + +我们只需派生 **ISubmixBufferListener** ，覆写`OnNewSubmixBuffer`处理音频数据，并将其注册到需要捕获 **FAudioDevice** 的 **USoundSubmix** 即可，这块代码依赖比较多，贴在文章中会比较乱，具体的实现大家可参考 **FNiagaraSubmixListener** 。 + +### DSP + +UE内置了一个非常完善的DSP模块—— **SignalProcessing** + +你几乎能在其中找到你想要的任何东西： + +![image-20240204204639645](Resources/image-20240204204639645.png) + +这是一些常用的结构和函数： + +- **Audio::TCircularAudioBuffer** ：环形缓冲区，常用于流式音频的存储。 +- **Audio::TSampleBuffer** ：用于将RawPCM数据转换为对应位深的可采样数据，类似于提供之前我们提供的`SampleValueWriteFunction`，不过只支持`int16`和`float`。 +- **Audio::FResampler** ：重采样器，用于修改音频的采样率。 +- **Audio::FWindow** ：用于生成窗函数，支持`Hamming`，`Hann`，`Blackman`。 +- **Audio::TSlidingWindow** ：滑动窗口。 +- **Audio::IFFTAlgorithm** ：提供基础的FFT算法，提供`FFFTFactory::NewFFTAlgorithm(...)`创建。 +- **Audio::FBiquadFilter** ：双二阶滤波器。 +- **Audio::FSpectrumAnalyzer** ：频谱分析仪，支持振幅谱，CQT谱。 +- **Audio::FAsyncSpectrumAnalyzer** ：执行在异步线程的频谱分析仪，功能同上。 +- `ConstantQ.h`：提供了一系列支持常量Q变换的结构，CQT常用于音乐的频谱可视化，它相较于FFT有更好的节奏表现效果。 + +对于频谱分析，UE还有一个模块`AudioSynesthesia`来扩展相关的功能。 + +![image-20240205104528872](Resources/image-20240205104528872.png) + +它支持 **CQT** ， **Onset（迸发）** ， **Loudness（响度）** ， **Meter** 的分析，可惜的是大多是一些 **NRT（非实时）** 接口，它要求音频数据必须是在编辑器时导入的资产，不过它里面很多算法结构可以供大家参考。 + +### MetaSound + +**Meta Sound** 是UE5推出的新一代音频系统，相较于旧版的 **Sound Cues** ， **Meta Sound** 不仅具有更好的性能， 更重要的是它支持 **程序化 ( Procedural )** + +目前 Meta Sound 在 **UGC** 方面 展现出了非常大的潜力： + +- **Mix Universe** ：[活动演讲 | 将音乐带入《Mix Universe》(官方字幕)](https://www.bilibili.com/video/BV1y84y1a7rr) +- **Fortnite** ：[堡垒之夜 - AR 编曲](https://www.bilibili.com/video/BV1F94y1L7fW) + +如果你还不了解 Meta Sound，强烈建议去看这个基础视频： + +- [活动演讲 | 初学者的MetaSound指南(官方字幕)](https://www.bilibili.com/video/BV15G4y1A7B5) + +新建一个 **MetaSoundSource** ，创建如下的图表，点击播放，你能听到熟悉的正弦波： + +![image-20240205111218033](Resources/image-20240205111218033.png) + +Meta Sound 的图表是一种数据流图，它的每个节点是一个功能函数块，它会接受Input，再结合MetaSound执行的当前上下文，任何填充Output数据。 + +以上面的正弦波为例，引擎在播放MetaSound时，会经过如下步骤： + +- 首先会创建构建MetaSound图表的异步任务，其对应的函数堆栈如下： + + ![image-20240205142028396](Resources/image-20240205142028396.png) + +- 在异步任务中，会加载所有节点（没有依赖的节点会被剔除），并根据依赖关系对剩下的节点进行拓扑排序，相应的代码堆栈如下： + + ![image-20240205143150293](Resources/image-20240205143150293.png) + +- 最终会通过节点内部的 **INodeOperatorFactory** 来创建 **IOperator** ， **IOperator** 的部分定义如下，它提供了`BindInput`，`ResetFunction`，`ExecuteFunction`，`PostExecuteFunction`来编写相关逻辑： + + ``` c++ + class IOperator + { + public: + /** FResetOperatorParams holds the parameters provided to an IOperator's + * reset function. + */ + struct FResetParams + { + /** General operator settings for the graph. */ + const FOperatorSettings& OperatorSettings; + + /** Environment settings available. */ + const FMetasoundEnvironment& Environment; + }; + + virtual ~IOperator() {} + + /** BindInputs binds data references in the IOperator with the FInputVertexInterfaceData. + * + * All input data references should be bound to the InVertexData to support + * other MetaSound systems such as MetaSound BP API, Operator Caching, + * and live auditioning. + * + * Note: The virtual function IOPerator::BindInputs(...) will be made a + * pure virtual when IOperator::GetInputs() is removed at or after release 5.5 + * + * Note: Binding an data reference may update the which underlying object + * the reference points to. Any operator which caches pointers or values + * from data references must update their cached pointers in the call to + * BindInputs(...). Operators which do not cache the underlying pointer + * of a data reference do not need to update anything after BindInputs(...) + * + * Example: + * FMyOperator::FMyOperator(TDataReadReference InGain, TDataReadReference InAudioBuffer) + * : Gain(InGain) + * , AudioBuffer(InAudioBuffer) + * { + * MyBufferPtr = AudioBuffer->GetData(); + * } + * + * void FMyOperator::BindInputs(FInputVertexInterfaceData& InVertexData) + * { + * InVertexData.BindReadVertex("Gain", Gain); + * InVertexData.BindReadVertex("Audio", AudioBuffer); + * + * // Update MyBufferPtr in case AudioBuffer references a new FAudioBuffer. + * MyBufferPtr = AudioBuffer->GetData(); + * } + */ + virtual void METASOUNDGRAPHCORE_API BindInputs(FInputVertexInterfaceData& InVertexData); + + /** BindOutputs binds data references in the IOperator with the FOutputVertexInterfaceData. + * + * All output data references should be bound to the InVertexData to support + * other MetaSound systems such as MetaSound BP API, Operator Caching, + * and live auditioning. + * + * Note: The virtual function IOPerator::BindOutputs(...) will be made a + * pure virtual when IOperator::GetOutputs() is removed at or after release 5.5 + * + * Note: Binding an data reference may update the which underlying object + * the reference points to. Any operator which caches pointers or values + * from data references must update their cached pointers in the call to + * BindOutputs(...). Operators which do not cache the underlying pointer + * of a data reference do not need to update anything after BindOutputs(...) + * + * Example: + * FMyOperator::FMyOperator(TDataWriteReference InGain, TDataWriteReference InAudioBuffer) + * : Gain(InGain) + * , AudioBuffer(InAudioBuffer) + * { + * MyBufferPtr = AudioBuffer->GetData(); + * } + * + * void FMyOperator::BindOutputs(FOutputVertexInterfaceData& InVertexData) + * { + * InVertexData.BindReadVertex("Gain", Gain); + * InVertexData.BindReadVertex("Audio", AudioBuffer); + * + * // Update MyBufferPtr in case AudioBuffer references a new FAudioBuffer. + * MyBufferPtr = AudioBuffer->GetData(); + * } + */ + virtual void METASOUNDGRAPHCORE_API BindOutputs(FOutputVertexInterfaceData& InVertexData); + + /** Pointer to initialize function for an operator. + * + * @param IOperator* - The operator associated with the function pointer. + */ + using FResetFunction = void(*)(IOperator*, const FResetParams& InParams); + + /** Return the reset function to call during graph execution. + * + * The IOperator* argument to the FExecutionFunction will be the same IOperator instance + * which returned the execution function. + * + * nullptr return values are valid and signal an IOperator which does not need to be + * reset. + */ + virtual FResetFunction GetResetFunction() = 0; + + /** Pointer to execute function for an operator. + * + * @param IOperator* - The operator associated with the function pointer. + */ + using FExecuteFunction = void(*)(IOperator*); + + /** Return the execution function to call during graph execution. + * + * The IOperator* argument to the FExecutionFunction will be the same IOperator instance + * which returned the execution function. + * + * nullptr return values are valid and signal an IOperator which does not need to be + * executed. + */ + virtual FExecuteFunction GetExecuteFunction() = 0; + + /** Pointer to post execute function for an operator. + * + * @param IOperator* - The operator associated with the function pointer. + */ + using FPostExecuteFunction = void(*)(IOperator*); + + /** Return the FPostExecute function to call during graph post execution. + * + * The IOperator* argument to the FPostExecutionFunction will be the same IOperator instance + * which returned the execution function. + * + * nullptr return values are valid and signal an IOperator which does not need to be + * post executed. + */ + virtual FPostExecuteFunction GetPostExecuteFunction() = 0; + }; + ``` + + ![image-20240205142303554](Resources/image-20240205142303554.png) + +- 经过上面的流程后，图表已经构建完毕，还记得我们之前提到过，如果 **USoundWave** 的`bProcedural`为`true`，它会创建一个相应的 **ISoundGenerator** ，并在音频引擎的每一帧，调用`ISoundGenerator::OnGenerateAudio`来填充数据，没错， **Meta Sound** 正是这样的流程： + + ![image-20240205143955462](Resources/image-20240205143955462.png) + +- 来看看我们的正弦节点在这一帧执行时做了什么： + + > 每一帧播放，它会根据`DeltaPhase`递增`Phase`，然后从 **FSineWaveTableGenerator** 的`WaveTable`中获取到相关的样本值，填充到输出的AudioBuffer中 + + ![image-20240205150518808](Resources/image-20240205150518808.png) + +![image-20240205150739219](Resources/image-20240205150739219.png) + +至此，相信你已经基本搞懂了 **Meta Sound** 的基本原理。 + +但要想用好Meta Sound，还需要更多专业的音频知识，这方面笔者也是一个菜鸡，有点爱莫能助~ + +它里面提供了很多的函数节点，试一试，电脑又不会爆炸： + +![image-20240205152403078](Resources/image-20240205152403078.png) + +但有时候并不能完全满足我们的需求，现在，你可以试着做一些自定义的扩展。 + +#### 蓝图交互 + +**Meta Sound ** 与蓝图并不相同，它支持的变量比较有限： + +![image-20240205152805417](Resources/image-20240205152805417.png) + +你同样也可以在C++中扩展这些变量，只需要找到合适的参考。 + +这些参数并没有直接暴露到属性编辑器，想要修改这些变量值有点复杂，我们可以借助 **UAudioComponent** 来简化这个过程，它派生自 **IAudioParameterControllerInterface** ，该接口提供了很多音频参数设置的便捷接口： + +![image-20240205154502428](Resources/image-20240205154502428.png) + +如果不想使用 **UAudioComponent** ，可以参考相关函数的具体实现。 + +这里需要特意提一下 UE 音频系统中的 `Trigger`，它与C++代码中常用的`Delegate`并不相同，我们可以简单把它当作一个bool变量，比如这样的图表： + +![image-20240205155420006](Resources/image-20240205155420006.png) + +已知每一个节点都会根据依赖关系的拓扑排序来执行，Input在执行时，会将输出的委托`播放时`置为true，在执行`AD Envelope`的时候，会读取输入委托`触发`与之相连的节点（即`播放时`），如果发现它为true才会执行该节点，在`AD Envelope`执行完毕后，会将输出委托`完成时`置为`true`，此时`Output`节点执行时检测到输入委托`完成时`为true，才会通知`MetaSoundSource`结束播放，也就是这里： + +![image-20240205155953096](Resources/image-20240205155953096.png) + +![image-20240205160115633](Resources/image-20240205160115633.png) + +#### 运行时构建 + +在 [Sounds Good](https://www.bilibili.com/video/BV1EC4y1J7Bs) 中，大佬提到了Meta Sound已经支持了运行时构建 Sound Graph，但是目前而言并不是很完善，还处于测试阶段，在 **MetasoundEngineTest** 中，能找到很多测试代码： + +![image-20240205160406177](Resources/image-20240205160406177.png) + +考虑到这部分代码后续可能会有较大的变动，因此暂不展开，后续官方版本稳定后再更新本节内容~ + +演示中还提到， Meta Sound 将推出大量新的音频控件！！！！！！！！！！！！ + +![image-20240205161259486](Resources/image-20240205161259486.png) + + + diff --git a/Docs/04-UnrealEngine/Resources/9796875423a0a9928413d91f8c7420c2.png b/Docs/04-UnrealEngine/Resources/9796875423a0a9928413d91f8c7420c2.png new file mode 100644 index 00000000..29348293 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/9796875423a0a9928413d91f8c7420c2.png differ diff --git a/Docs/04-UnrealEngine/Resources/OIP-C.jpeg b/Docs/04-UnrealEngine/Resources/OIP-C.jpeg new file mode 100644 index 00000000..08ae1f0c Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/OIP-C.jpeg differ diff --git a/Docs/04-UnrealEngine/Resources/R-C.2b237c95a4c38cc9268190c0929c099arik=jMO03ZbheSf1Iw&riu=http%3a%2f%2fwww.rlocman.ru%2fi%2fImage%2f2009%2f07%2f03%2f1-1707012302768-7.png&ehk=WtHxNxcRmjPsoTfOhVAzqLifzBQfHjSQt9lqsI71n%2f8%3d&risl=&pid=ImgRaw&r=0&sres=1&sresct=1 b/Docs/04-UnrealEngine/Resources/R-C.2b237c95a4c38cc9268190c0929c099arik=jMO03ZbheSf1Iw&riu=http%3a%2f%2fwww.rlocman.ru%2fi%2fImage%2f2009%2f07%2f03%2f1-1707012302768-7.png&ehk=WtHxNxcRmjPsoTfOhVAzqLifzBQfHjSQt9lqsI71n%2f8%3d&risl=&pid=ImgRaw&r=0&sres=1&sresct=1 new file mode 100644 index 00000000..e3fbec70 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/R-C.2b237c95a4c38cc9268190c0929c099arik=jMO03ZbheSf1Iw&riu=http%3a%2f%2fwww.rlocman.ru%2fi%2fImage%2f2009%2f07%2f03%2f1-1707012302768-7.png&ehk=WtHxNxcRmjPsoTfOhVAzqLifzBQfHjSQt9lqsI71n%2f8%3d&risl=&pid=ImgRaw&r=0&sres=1&sresct=1 differ diff --git a/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011297990-1.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 b/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011297990-1.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 new file mode 100644 index 00000000..e6a7c7fe Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011297990-1.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 differ diff --git a/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011329329-2.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 b/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011329329-2.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 new file mode 100644 index 00000000..ad3eb0c6 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011329329-2.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 differ diff --git a/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011331880-3.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 b/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011331880-3.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 new file mode 100644 index 00000000..b41e0ec8 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/R-C.a243ac1e5f3787cc0496bb865fd1febcrik=ZTMdoh9CEIIdMQ&riu=http%3a%2f%2fwww.stgeorgeutah.com%2fwp-content%2fuploads%2f2015%2f10%2fDollarphotoclub_88754239-1707011331880-3.jpg&ehk=DmZ91TTNfITQFOeBcJAz%2be5WvrCDTvI%2b2AcKM3D9TOI%3d&risl=&pid=ImgRaw&r=0 differ diff --git a/Docs/04-UnrealEngine/Resources/The-structure-of-wav-file-format.png b/Docs/04-UnrealEngine/Resources/The-structure-of-wav-file-format.png new file mode 100644 index 00000000..e39bc185 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/The-structure-of-wav-file-format.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204102619133.png b/Docs/04-UnrealEngine/Resources/image-20240204102619133.png new file mode 100644 index 00000000..76ab8008 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204102619133.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204103616957.png b/Docs/04-UnrealEngine/Resources/image-20240204103616957.png new file mode 100644 index 00000000..d09c8e9e Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204103616957.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204104205505.png b/Docs/04-UnrealEngine/Resources/image-20240204104205505.png new file mode 100644 index 00000000..4a6f3a6c Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204104205505.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204105850144.png b/Docs/04-UnrealEngine/Resources/image-20240204105850144.png new file mode 100644 index 00000000..9ac19343 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204105850144.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204141548048.png b/Docs/04-UnrealEngine/Resources/image-20240204141548048.png new file mode 100644 index 00000000..23c0ff58 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204141548048.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204142046638.png b/Docs/04-UnrealEngine/Resources/image-20240204142046638.png new file mode 100644 index 00000000..573ab8ea Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204142046638.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204142115830.png b/Docs/04-UnrealEngine/Resources/image-20240204142115830.png new file mode 100644 index 00000000..45c3341f Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204142115830.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204155736479.png b/Docs/04-UnrealEngine/Resources/image-20240204155736479.png new file mode 100644 index 00000000..540a212d Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204155736479.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204170415081.png b/Docs/04-UnrealEngine/Resources/image-20240204170415081.png new file mode 100644 index 00000000..e84f0b4b Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204170415081.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204191436270.png b/Docs/04-UnrealEngine/Resources/image-20240204191436270.png new file mode 100644 index 00000000..d9a07e49 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204191436270.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204191454472.png b/Docs/04-UnrealEngine/Resources/image-20240204191454472.png new file mode 100644 index 00000000..a52ee427 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204191454472.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204195219544.png b/Docs/04-UnrealEngine/Resources/image-20240204195219544.png new file mode 100644 index 00000000..cf89d25f Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204195219544.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204195409513.png b/Docs/04-UnrealEngine/Resources/image-20240204195409513.png new file mode 100644 index 00000000..763ef070 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204195409513.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240204204639645.png b/Docs/04-UnrealEngine/Resources/image-20240204204639645.png new file mode 100644 index 00000000..804794fc Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240204204639645.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205104528872.png b/Docs/04-UnrealEngine/Resources/image-20240205104528872.png new file mode 100644 index 00000000..8174876a Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205104528872.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205111218033.png b/Docs/04-UnrealEngine/Resources/image-20240205111218033.png new file mode 100644 index 00000000..6d58c738 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205111218033.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205111652587.png b/Docs/04-UnrealEngine/Resources/image-20240205111652587.png new file mode 100644 index 00000000..ac37b546 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205111652587.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205142028396.png b/Docs/04-UnrealEngine/Resources/image-20240205142028396.png new file mode 100644 index 00000000..05f97ed1 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205142028396.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205142303554.png b/Docs/04-UnrealEngine/Resources/image-20240205142303554.png new file mode 100644 index 00000000..82e19eea Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205142303554.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205143150293.png b/Docs/04-UnrealEngine/Resources/image-20240205143150293.png new file mode 100644 index 00000000..fe47a499 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205143150293.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205143955462.png b/Docs/04-UnrealEngine/Resources/image-20240205143955462.png new file mode 100644 index 00000000..09bc540d Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205143955462.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205144327362.png b/Docs/04-UnrealEngine/Resources/image-20240205144327362.png new file mode 100644 index 00000000..eb7ee981 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205144327362.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205150518808.png b/Docs/04-UnrealEngine/Resources/image-20240205150518808.png new file mode 100644 index 00000000..60770525 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205150518808.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205150739219.png b/Docs/04-UnrealEngine/Resources/image-20240205150739219.png new file mode 100644 index 00000000..6e8ffaea Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205150739219.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205152403078.png b/Docs/04-UnrealEngine/Resources/image-20240205152403078.png new file mode 100644 index 00000000..a2d8bfee Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205152403078.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205152805417.png b/Docs/04-UnrealEngine/Resources/image-20240205152805417.png new file mode 100644 index 00000000..4514db6d Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205152805417.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205154502428.png b/Docs/04-UnrealEngine/Resources/image-20240205154502428.png new file mode 100644 index 00000000..45c1b996 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205154502428.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205155420006.png b/Docs/04-UnrealEngine/Resources/image-20240205155420006.png new file mode 100644 index 00000000..b3fbfae1 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205155420006.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205155953096.png b/Docs/04-UnrealEngine/Resources/image-20240205155953096.png new file mode 100644 index 00000000..6667348b Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205155953096.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205160115633.png b/Docs/04-UnrealEngine/Resources/image-20240205160115633.png new file mode 100644 index 00000000..9f0573bf Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205160115633.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205160406177.png b/Docs/04-UnrealEngine/Resources/image-20240205160406177.png new file mode 100644 index 00000000..4ed5d4e1 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205160406177.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205161259486.png b/Docs/04-UnrealEngine/Resources/image-20240205161259486.png new file mode 100644 index 00000000..bf5019cf Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205161259486.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205162426317.png b/Docs/04-UnrealEngine/Resources/image-20240205162426317.png new file mode 100644 index 00000000..87202e4f Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205162426317.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205162859891.png b/Docs/04-UnrealEngine/Resources/image-20240205162859891.png new file mode 100644 index 00000000..076ea8d3 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205162859891.png differ diff --git a/Docs/04-UnrealEngine/Resources/image-20240205162920488.png b/Docs/04-UnrealEngine/Resources/image-20240205162920488.png new file mode 100644 index 00000000..2f190442 Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/image-20240205162920488.png differ diff --git a/Docs/04-UnrealEngine/Resources/sfsa.gif b/Docs/04-UnrealEngine/Resources/sfsa.gif new file mode 100644 index 00000000..24b6311b Binary files /dev/null and b/Docs/04-UnrealEngine/Resources/sfsa.gif differ