我可以使用 AVAudioEngine 从文件中读取，使用音频单元处理并写入文件，比实时更快吗？

Question

我正在开发一个 iOS 应用程序，该应用程序使用 AVAudioEngine 进行各种操作，包括将音频录制到文件、使用音频单元对该音频应用效果以及播放应用了效果的音频。 我还使用水龙头将输出写入文件。 完成此操作后，它会在播放音频时实时写入文件。

是否可以设置一个 AVAudioEngine 图从文件中读取，用音频单元处理声音，然后输出到文件，但比实时更快（即，硬件处理它的速度快）？ 这样做的用例是输出几分钟应用了效果的音频，我当然不想等待几分钟来处理它。

编辑：这是我用来设置 AVAudioEngine 图形并播放声音文件的代码：

AVAudioEngine* engine = [[AVAudioEngine alloc] init];

AVAudioPlayerNode* player = [[AVAudioPlayerNode alloc] init];
[engine attachNode:player];

self.player = player;
self.engine = engine;

if (!self.distortionEffect) {
    self.distortionEffect = [[AVAudioUnitDistortion alloc] init];
    [self.engine attachNode:self.distortionEffect];
    [self.engine connect:self.player to:self.distortionEffect format:[self.distortionEffect outputFormatForBus:0]];
    AVAudioMixerNode* mixer = [self.engine mainMixerNode];
    [self.engine connect:self.distortionEffect to:mixer format:[mixer outputFormatForBus:0]];
}

[self.distortionEffect loadFactoryPreset:AVAudioUnitDistortionPresetDrumsBitBrush];

NSError* error;
if (![self.engine startAndReturnError:&error]) {
    NSLog(@"error: %@", error);
} else {
    NSURL* fileURL = [[NSBundle mainBundle] URLForResource:@"test2" withExtension:@"mp3"];
    AVAudioFile* file = [[AVAudioFile alloc] initForReading:fileURL error:&error];

    if (error) {
        NSLog(@"error: %@", error);
    } else {
        [self.player scheduleFile:file atTime:nil completionHandler:nil];
        [self.player play];
    }
}

上面的代码实时播放 test2.mp3 文件中的声音，并应用了AVAudioUnitDistortionPresetDrumsBitBrush失真预设。

然后我通过在 [self.player play] 之后添加这些行来修改上面的代码：

        [self.engine stop];
        [self renderAudioAndWriteToFile];

我修改了 Vladimir 提供的 renderAudioAndWriteToFile 方法，这样它就不会在第一行分配一个新的 AVAudioEngine，而是使用已经设置好的 self.engine。

但是，在 renderAudioAndWriteToFile 中，它记录“无法呈现音频单元”，因为 AudioUnitRender 正在返回kAudioUnitErr_Uninitialized状态。

编辑 2 ：我应该提到我非常乐意将我发布的 AVAudioEngine 代码转换为使用 C api，如果这会让事情变得更容易。 但是，我希望代码产生与 AVAudioEngine 代码相同的输出（包括使用上面显示的工厂预设）。

Answer 1

1. 配置您的引擎和播放器节点。
2. 为您的播放器节点调用play方法。
3. 暂停你的引擎。
4. 使用此方法从 AVAudioOutputNode ( audioEngine.outputNode ) 获取音频单元。
5. 使用AudioUnitRender循环从音频单元渲染，并使用扩展音频文件服务将音频缓冲区列表写入文件。

例子：

音频引擎配置

- (void)configureAudioEngine {
    self.engine = [[AVAudioEngine alloc] init];
    self.playerNode = [[AVAudioPlayerNode alloc] init];
    [self.engine attachNode:self.playerNode];
    AVAudioUnitDistortion *distortionEffect = [[AVAudioUnitDistortion alloc] init];
    [self.engine attachNode:distortionEffect];
    [self.engine connect:self.playerNode to:distortionEffect format:[distortionEffect outputFormatForBus:0]];
    self.mixer = [self.engine mainMixerNode];
    [self.engine connect:distortionEffect to:self.mixer format:[self.mixer outputFormatForBus:0]];
    [distortionEffect loadFactoryPreset:AVAudioUnitDistortionPresetDrumsBitBrush];
    NSError* error;
    if (![self.engine startAndReturnError:&error])
        NSLog(@"Can't start engine: %@", error);
    else
        [self scheduleFileToPlay];
}

- (void)scheduleFileToPlay {
    NSError* error;
    NSURL *fileURL = [[NSBundle mainBundle] URLForResource:@"filename" withExtension:@"m4a"];
    self.file = [[AVAudioFile alloc] initForReading:fileURL error:&error];
    if (self.file)
        [self.playerNode scheduleFile:self.file atTime:nil completionHandler:nil];
    else
        NSLog(@"Can't read file: %@", error);
}

渲染方法

- (void)renderAudioAndWriteToFile {
    [self.playerNode play];
    [self.engine pause];
    AVAudioOutputNode *outputNode = self.engine.outputNode;
    AudioStreamBasicDescription const *audioDescription = [outputNode outputFormatForBus:0].streamDescription;
    NSString *path = [self filePath];
    ExtAudioFileRef audioFile = [self createAndSetupExtAudioFileWithASBD:audioDescription andFilePath:path];
    if (!audioFile)
        return;
    AVURLAsset *asset = [AVURLAsset assetWithURL:self.file.url];
    NSTimeInterval duration = CMTimeGetSeconds(asset.duration);
    NSUInteger lengthInFrames = duration * audioDescription->mSampleRate;
    const NSUInteger kBufferLength = 4096;
    AudioBufferList *bufferList = AEAllocateAndInitAudioBufferList(*audioDescription, kBufferLength);
    AudioTimeStamp timeStamp;
    memset (&timeStamp, 0, sizeof(timeStamp));
    timeStamp.mFlags = kAudioTimeStampSampleTimeValid;
    OSStatus status = noErr;
    for (NSUInteger i = kBufferLength; i < lengthInFrames; i += kBufferLength) {
        status = [self renderToBufferList:bufferList writeToFile:audioFile bufferLength:kBufferLength timeStamp:&timeStamp];
        if (status != noErr)
            break;
    }
    if (status == noErr && timeStamp.mSampleTime < lengthInFrames) {
        NSUInteger restBufferLength = (NSUInteger) (lengthInFrames - timeStamp.mSampleTime);
        AudioBufferList *restBufferList = AEAllocateAndInitAudioBufferList(*audioDescription, restBufferLength);
        status = [self renderToBufferList:restBufferList writeToFile:audioFile bufferLength:restBufferLength timeStamp:&timeStamp];
        AEFreeAudioBufferList(restBufferList);
    }
    AEFreeAudioBufferList(bufferList);
    ExtAudioFileDispose(audioFile);
    if (status != noErr)
        NSLog(@"An error has occurred");
    else
        NSLog(@"Finished writing to file at path: %@", path);
}

- (NSString *)filePath {
    NSArray *documentsFolders =
            NSSearchPathForDirectoriesInDomains(NSDocumentDirectory, NSUserDomainMask, YES);
    NSString *fileName = [NSString stringWithFormat:@"%@.m4a", [[NSUUID UUID] UUIDString]];
    NSString *path = [documentsFolders[0] stringByAppendingPathComponent:fileName];
    return path;
}

- (ExtAudioFileRef)createAndSetupExtAudioFileWithASBD:(AudioStreamBasicDescription const *)audioDescription
                                          andFilePath:(NSString *)path {
    AudioStreamBasicDescription destinationFormat;
    memset(&destinationFormat, 0, sizeof(destinationFormat));
    destinationFormat.mChannelsPerFrame = audioDescription->mChannelsPerFrame;
    destinationFormat.mSampleRate = audioDescription->mSampleRate;
    destinationFormat.mFormatID = kAudioFormatMPEG4AAC;
    ExtAudioFileRef audioFile;
    OSStatus status = ExtAudioFileCreateWithURL(
            (__bridge CFURLRef) [NSURL fileURLWithPath:path],
            kAudioFileM4AType,
            &destinationFormat,
            NULL,
            kAudioFileFlags_EraseFile,
            &audioFile
    );
    if (status != noErr) {
        NSLog(@"Can not create ext audio file");
        return nil;
    }
    UInt32 codecManufacturer = kAppleSoftwareAudioCodecManufacturer;
    status = ExtAudioFileSetProperty(
            audioFile, kExtAudioFileProperty_CodecManufacturer, sizeof(UInt32), &codecManufacturer
    );
    status = ExtAudioFileSetProperty(
            audioFile, kExtAudioFileProperty_ClientDataFormat, sizeof(AudioStreamBasicDescription), audioDescription
    );
    status = ExtAudioFileWriteAsync(audioFile, 0, NULL);
    if (status != noErr) {
        NSLog(@"Can not setup ext audio file");
        return nil;
    }
    return audioFile;
}

- (OSStatus)renderToBufferList:(AudioBufferList *)bufferList
                   writeToFile:(ExtAudioFileRef)audioFile
                  bufferLength:(NSUInteger)bufferLength
                     timeStamp:(AudioTimeStamp *)timeStamp {
    [self clearBufferList:bufferList];
    AudioUnit outputUnit = self.engine.outputNode.audioUnit;
    OSStatus status = AudioUnitRender(outputUnit, 0, timeStamp, 0, bufferLength, bufferList);
    if (status != noErr) {
        NSLog(@"Can not render audio unit");
        return status;
    }
    timeStamp->mSampleTime += bufferLength;
    status = ExtAudioFileWrite(audioFile, bufferLength, bufferList);
    if (status != noErr)
        NSLog(@"Can not write audio to file");
    return status;
}

- (void)clearBufferList:(AudioBufferList *)bufferList {
    for (int bufferIndex = 0; bufferIndex < bufferList->mNumberBuffers; bufferIndex++) {
        memset(bufferList->mBuffers[bufferIndex].mData, 0, bufferList->mBuffers[bufferIndex].mDataByteSize);
    }
}

我使用了这个很酷的框架中的一些函数：

AudioBufferList *AEAllocateAndInitAudioBufferList(AudioStreamBasicDescription audioFormat, int frameCount) {
    int numberOfBuffers = audioFormat.mFormatFlags & kAudioFormatFlagIsNonInterleaved ? audioFormat.mChannelsPerFrame : 1;
    int channelsPerBuffer = audioFormat.mFormatFlags & kAudioFormatFlagIsNonInterleaved ? 1 : audioFormat.mChannelsPerFrame;
    int bytesPerBuffer = audioFormat.mBytesPerFrame * frameCount;
    AudioBufferList *audio = malloc(sizeof(AudioBufferList) + (numberOfBuffers-1)*sizeof(AudioBuffer));
    if ( !audio ) {
        return NULL;
    }
    audio->mNumberBuffers = numberOfBuffers;
    for ( int i=0; i<numberOfBuffers; i++ ) {
        if ( bytesPerBuffer > 0 ) {
            audio->mBuffers[i].mData = calloc(bytesPerBuffer, 1);
            if ( !audio->mBuffers[i].mData ) {
                for ( int j=0; j<i; j++ ) free(audio->mBuffers[j].mData);
                free(audio);
                return NULL;
            }
        } else {
            audio->mBuffers[i].mData = NULL;
        }
        audio->mBuffers[i].mDataByteSize = bytesPerBuffer;
        audio->mBuffers[i].mNumberChannels = channelsPerBuffer;
    }
    return audio;
}

void AEFreeAudioBufferList(AudioBufferList *bufferList ) {
    for ( int i=0; i<bufferList->mNumberBuffers; i++ ) {
        if ( bufferList->mBuffers[i].mData ) free(bufferList->mBuffers[i].mData);
    }
    free(bufferList);
}

Answer 2

对于如今正在寻找此问题答案的任何人来说，Apple 似乎只是为此目的添加了离线处理功能： 请参阅文档。