如果您想自己运行此项目,则这是一个完整的项目:https://www.dropbox.com/s/5p384mogjzflvqk/AVPlayerLayerSoundOnlyBug_iOS10.zip?dl=0
这是iOS 10上的一个新问题,自iOS 10.2起已得到修复。在使用AVAssetExportSession和AVVideoCompositionCoreAnimationTool导出视频以在导出过程中在视频之上合成一个图层后,在AVPlayerLayer中播放的视频将无法播放。这似乎不是由达到AV编码/解码管道限制引起的,因为它通常在单个导出之后发生,据我所知,这仅增加了2个管道:1个用于AVAssetExportSession,另一个用于AVPlayer。我还正确设置了图层的框架,如下面的代码所示,您可以通过运行下面的代码看到该图层,使该图层可以清楚地看到蓝色背景。
导出后,在播放视频之前等待一些时间似乎可以使它更加可靠,但这并不是告诉用户的真正可接受的解决方法。
关于是什么原因引起的任何想法,或者如何解决或解决它的任何想法?我是否弄乱了某些内容或错过了重要的步骤或细节?任何帮助或指向文档的指针都将不胜感激。
import UIKit
import AVFoundation
/* After exporting an AVAsset using AVAssetExportSession with AVVideoCompositionCoreAnimationTool, we
* will attempt to play a video using an AVPlayerLayer with a blue background.
*
* If you see the blue background and hear audio you're experiencing the missing-video bug. Otherwise
* try hitting the button again.
*/
class ViewController: UIViewController {
private var playerLayer: AVPlayerLayer?
private let button = UIButton()
private let indicator = UIActivityIndicatorView(activityIndicatorStyle: .gray)
override func viewDidLoad() {
super.viewDidLoad()
view.backgroundColor = UIColor.white
button.setTitle("Cause Trouble", for: .normal)
button.setTitleColor(UIColor.black, for: .normal)
button.addTarget(self, action: #selector(ViewController.buttonTapped), for: .touchUpInside)
view.addSubview(button)
button.translatesAutoresizingMaskIntoConstraints = false
NSLayoutConstraint.activate([
button.centerXAnchor.constraint(equalTo: view.centerXAnchor),
button.bottomAnchor.constraint(equalTo: view.bottomAnchor, constant: -16),
])
indicator.hidesWhenStopped = true
view.insertSubview(indicator, belowSubview: button)
indicator.translatesAutoresizingMaskIntoConstraints = false
NSLayoutConstraint.activate([
indicator.centerXAnchor.constraint(equalTo: button.centerXAnchor),
indicator.centerYAnchor.constraint(equalTo: button.centerYAnchor),
])
}
func buttonTapped() {
button.isHidden = true
indicator.startAnimating()
playerLayer?.removeFromSuperlayer()
let sourcePath = Bundle.main.path(forResource: "video.mov", ofType: nil)!
let sourceURL = URL(fileURLWithPath: sourcePath)
let sourceAsset = AVURLAsset(url: sourceURL)
//////////////////////////////////////////////////////////////////////
// STEP 1: Export a video using AVVideoCompositionCoreAnimationTool //
//////////////////////////////////////////////////////////////////////
let exportSession = { () -> AVAssetExportSession in
let sourceTrack = sourceAsset.tracks(withMediaType: AVMediaTypeVideo).first!
let parentLayer = CALayer()
parentLayer.frame = CGRect(origin: .zero, size: CGSize(width: 1280, height: 720))
let videoLayer = CALayer()
videoLayer.frame = parentLayer.bounds
parentLayer.addSublayer(videoLayer)
let composition = AVMutableVideoComposition(propertiesOf: sourceAsset)
composition.animationTool = AVVideoCompositionCoreAnimationTool(postProcessingAsVideoLayer: videoLayer, in: parentLayer)
let layerInstruction = AVMutableVideoCompositionLayerInstruction(assetTrack: sourceTrack)
layerInstruction.setTransform(sourceTrack.preferredTransform, at: kCMTimeZero)
let instruction = AVMutableVideoCompositionInstruction()
instruction.timeRange = CMTimeRange(start: kCMTimeZero, duration: sourceAsset.duration)
instruction.layerInstructions = [layerInstruction]
composition.instructions = [instruction]
let e = AVAssetExportSession(asset: sourceAsset, presetName: AVAssetExportPreset1280x720)!
e.videoComposition = composition
e.outputFileType = AVFileTypeQuickTimeMovie
e.timeRange = CMTimeRange(start: kCMTimeZero, duration: sourceAsset.duration)
let outputURL = URL(fileURLWithPath: NSTemporaryDirectory().appending("/out2.mov"))
_ = try? FileManager.default.removeItem(at: outputURL)
e.outputURL = outputURL
return e
}()
print("Exporting asset...")
exportSession.exportAsynchronously {
assert(exportSession.status == .completed)
//////////////////////////////////////////////
// STEP 2: Play a video in an AVPlayerLayer //
//////////////////////////////////////////////
DispatchQueue.main.async {
// Reuse player layer, shouldn't be hitting the AV pipeline limit
let playerItem = AVPlayerItem(asset: sourceAsset)
let layer = self.playerLayer ?? AVPlayerLayer()
if layer.player == nil {
layer.player = AVPlayer(playerItem: playerItem)
}
else {
layer.player?.replaceCurrentItem(with: playerItem)
}
layer.backgroundColor = UIColor.blue.cgColor
if UIDeviceOrientationIsPortrait(UIDevice.current.orientation) {
layer.frame = self.view.bounds
layer.bounds.size.height = layer.bounds.width * 9.0 / 16.0
}
else {
layer.frame = self.view.bounds.insetBy(dx: 0, dy: 60)
layer.bounds.size.width = layer.bounds.height * 16.0 / 9.0
}
self.view.layer.insertSublayer(layer, at: 0)
self.playerLayer = layer
layer.player?.play()
print("Playing a video in an AVPlayerLayer...")
self.button.isHidden = false
self.indicator.stopAnimating()
}
}
}
}
最佳答案
在这种情况下,我的答案是通过使用实现AVVideoCompositionCoreAnimationTool
协议(protocol)的自定义视频合成类和实现AVVideoCompositing
协议(protocol)的自定义合成指令来解决AVVideoCompositionInstruction
的问题。因为我需要在视频顶部叠加CALayer
,所以我将这一层包括在合成指令实例中。
您需要像这样设置视频合成中的自定义合成器:
composition.customVideoCompositorClass = CustomVideoCompositor.self
然后在其上设置您的自定义说明:
let instruction = CustomVideoCompositionInstruction(...) // whatever parameters you need and are required by the instruction protocol
composition.instructions = [instruction]
编辑:这是一个如何使用自定义合成器使用GPU在视频上覆盖图层的工作示例:https://github.com/samsonjs/LayerVideoCompositor ...原始答案在下面继续
至于合成器本身,如果您观看相关的WWDC session 并查看其示例代码,则可以实现一个合成器。我无法在此处发布我写的内容,但是我正在使用CoreImage在处理
AVAsynchronousVideoCompositionRequest
方面进行繁重的工作,请确保使用OpenGL CoreImage上下文以获得最佳性能(如果在CPU上执行它将非常慢)。如果在导出过程中出现内存使用高峰,则可能还需要一个自动释放池。如果您像我一样覆盖
CALayer
,那么在将图层渲染为layer.isGeometryFlipped = true
之前,请确保设置CGImage
,然后再将其发送到CoreImage。并确保在合成器中逐帧缓存呈现的CGImage
。