这是汤米答案的快速版本。

 // Set up the Capture Session 
 // Add the Inputs 
 // Add the Outputs


 var outputSettings = [
    AVVideoWidthKey : Int(640),
    AVVideoHeightKey : Int(480),
    AVVideoCodecKey : .h264
]

    var assetWriterInput = AVAssetWriterInput(mediaType: AVMediaTypeVideo,outputSettings: outputSettings)

    var pixelBufferAdaptor = AVAssetWriterInputPixelBufferAdaptor(assetWriterInput, sourcePixelBufferAttributes:
        [ kCVPixelBufferPixelFormatTypeKey : Int(kCVPixelFormatType_32BGRA)])


     var assetWriter = AVAssetWriter(url: URLFromSomwhere, fileType: AVFileTypeMPEG4 , error : Error )
         assetWriter.addInput(assetWriterInput)
         assetWriterInput.expectsMediaDataInRealTime = true
         assetWriter.startWriting()
         assetWriter.startSession(atSourceTime: kCMTimeZero)

    captureSession.startRunning()


  func captureOutput(_ captureOutput: AVCaptureOutput, didOutputSampleBuffer sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {

    var imageBuffer = CMSampleBufferGetImageBuffer(sampleBuffer)
    // a very dense way to keep track of the time at which this frame
    // occurs relative to the output stream, but it's just an example!
    var frameNumber: Int64 = 0

           if assetWriterInput.readyForMoreMediaData {
    pixelBufferAdaptor.appendPixelBuffer(imageBuffer, withPresentationTime: CMTimeMake(frameNumber, 25))
               }
                frameNumber += 1   }

      captureSession.stopRunning()
      assetWriter.finishWriting()

我不保证100％的准确性，因为我是新手。

我无法回答提出的具体问题，但是我已经成功地使用以下方法同时录制了视频并抓取帧：

• AVCaptureSession和AVCaptureVideoDataOutput将帧路由到我自己的代码中
• AVAssetWriter ， AVAssetWriterInput和AVAssetWriterInputPixelBufferAdaptor将帧写出到H.264编码的电影文件中

无需调查音频。我最终从捕获会话中获取CMSampleBuffers ，然后将其推入像素缓冲区适配器。

编辑：所以我的代码或多或少看起来像，您可以略过并忽略范围的问题：

/* to ensure I'm given incoming CMSampleBuffers */
AVCaptureSession *captureSession = alloc and init, set your preferred preset/etc;
AVCaptureDevice *captureDevice = default for video, probably;

AVCaptureDeviceInput *deviceInput = input with device as above, 
                                    and attach it to the session;

AVCaptureVideoDataOutput *output = output for 32BGRA pixel format, with me as the
                                   delegate and a suitable dispatch queue affixed.

/* to prepare for output; I'll output 640x480 in H.264, via an asset writer */
NSDictionary *outputSettings =
    [NSDictionary dictionaryWithObjectsAndKeys:

            [NSNumber numberWithInt:640], AVVideoWidthKey,
            [NSNumber numberWithInt:480], AVVideoHeightKey,
            AVVideoCodecH264, AVVideoCodecKey,

            nil];

AVAssetWriterInput *assetWriterInput = [AVAssetWriterInput 
                                   assetWriterInputWithMediaType:AVMediaTypeVideo
                                                  outputSettings:outputSettings];

/* I'm going to push pixel buffers to it, so will need a 
   AVAssetWriterPixelBufferAdaptor, to expect the same 32BGRA input as I've
   asked the AVCaptureVideDataOutput to supply */
AVAssetWriterInputPixelBufferAdaptor *pixelBufferAdaptor =
           [[AVAssetWriterInputPixelBufferAdaptor alloc] 
                initWithAssetWriterInput:assetWriterInput 
                sourcePixelBufferAttributes:
                     [NSDictionary dictionaryWithObjectsAndKeys:
                          [NSNumber numberWithInt:kCVPixelFormatType_32BGRA], 
                           kCVPixelBufferPixelFormatTypeKey,
                     nil]];

/* that's going to go somewhere, I imagine you've got the URL for that sorted,
   so create a suitable asset writer; we'll put our H.264 within the normal
   MPEG4 container */
AVAssetWriter *assetWriter = [[AVAssetWriter alloc]
                                initWithURL:URLFromSomwhere
                                fileType:AVFileTypeMPEG4
                                error:you need to check error conditions,
                                      this example is too lazy];
[assetWriter addInput:assetWriterInput];

/* we need to warn the input to expect real time data incoming, so that it tries
   to avoid being unavailable at inopportune moments */
assetWriterInput.expectsMediaDataInRealTime = YES;

... eventually ...

[assetWriter startWriting];
[assetWriter startSessionAtSourceTime:kCMTimeZero];
[captureSession startRunning];

... elsewhere ...

- (void)        captureOutput:(AVCaptureOutput *)captureOutput 
    didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer 
           fromConnection:(AVCaptureConnection *)connection
{
    CVImageBufferRef imageBuffer = CMSampleBufferGetImageBuffer(sampleBuffer);

    // a very dense way to keep track of the time at which this frame
    // occurs relative to the output stream, but it's just an example!
    static int64_t frameNumber = 0;
    if(assetWriterInput.readyForMoreMediaData)
        [pixelBufferAdaptor appendPixelBuffer:imageBuffer
                         withPresentationTime:CMTimeMake(frameNumber, 25)];
    frameNumber++;
}

... and, to stop, ensuring the output file is finished properly ...

[captureSession stopRunning];
[assetWriter finishWriting];