Adding watermark to video is extremely slow - ios

I am using AVComposition to render a watermark to a video. This process takes around 15 seconds, which doesn't seem ok for a 20 seconds video.
My export settings are:
let exporter = AVAssetExportSession.init(asset: mixComposition, presetName: AVAssetExportPresetHighestQuality)
exporter?.outputURL = outputPath
exporter?.outputFileType = AVFileType.mp4
exporter?.shouldOptimizeForNetworkUse = true
exporter?.videoComposition = mainCompositionInst
DispatchQueue.main.async {
exporter?.exportAsynchronously(completionHandler: {
if exporter?.status == AVAssetExportSessionStatus.completed {
completion(true, exporter)
}else{
completion(false, exporter)
}
})
}
This is how I add the watermark:
//Creating image layer
let overlayLayer = CALayer()
let overlayImage: UIImage = image
overlayLayer.contents = overlayImage.cgImage
overlayLayer.frame = CGRect(x: 0, y: 0, width: size.width, height: size.height)
overlayLayer.contentsGravity = kCAGravityResizeAspectFill
overlayLayer.masksToBounds = true
//Creating parent and video layer
let parentLayer = CALayer()
let videoLayer = CALayer()
parentLayer.frame = CGRect(x: 0, y: 0, width: size.width, height: size.height)
videoLayer.frame = CGRect(x: 0, y: 0, width: size.width, height: size.height)
parentLayer.addSublayer(videoLayer)
parentLayer.addSublayer(overlayLayer)
//Adding those layers to video
composition.animationTool = AVVideoCompositionCoreAnimationTool(postProcessingAsVideoLayer: videoLayer, in: parentLayer)
}
and this is how I eventually transform my video:
let videoLayerInstruction = AVMutableVideoCompositionLayerInstruction.init(assetTrack: videoTrack!)
let videoAssetTrack = videoAsset.tracks(withMediaType: AVMediaType.video)[0]
var videoAssetOrientation = UIImageOrientation.up
var isVideoAssetPortrait = false
let videoTransform = videoAssetTrack.preferredTransform
if videoTransform.a == 0 && videoTransform.b == 1.0 && videoTransform.c == -1.0 && videoTransform.d == 0 {
videoAssetOrientation = .right
isVideoAssetPortrait = true
}
if videoTransform.a == 0 && videoTransform.b == -1.0 && videoTransform.c == 1.0 && videoTransform.d == 0 {
videoAssetOrientation = .left
isVideoAssetPortrait = true
}
if videoTransform.a == 1.0 && videoTransform.b == 0 && videoTransform.c == 0 && videoTransform.d == 1.0 {
videoAssetOrientation = .up
}
if videoTransform.a == -1.0 && videoTransform.b == 0 && videoTransform.c == 0 && videoTransform.d == -1.0 {
videoAssetOrientation = .down
}
videoLayerInstruction.setTransform(videoAssetTrack.preferredTransform, at: kCMTimeZero)
//Add instructions
mainInstruction.layerInstructions = [videoLayerInstruction]
let mainCompositionInst = AVMutableVideoComposition()
let naturalSize : CGSize!
if isVideoAssetPortrait {
naturalSize = CGSize(width: videoAssetTrack.naturalSize.height, height: videoAssetTrack.naturalSize.width)
} else {
naturalSize = videoAssetTrack.naturalSize
}
So my question now is, how can I improve the performance of merging the watermark to my video? 15 seconds is totally unacceptable for any kind of end-user. Furthermore, I need to transport this video over the internet, so the loading screen would show its beauty more than a total of approximately twenty seconds.

Per the Apple Documentation, try using the class AVAsynchronousCIImageFilteringRequest
Overview
You use this class when creating a composition for Core Image filtering with the init(asset:applyingCIFiltersWithHandler:) method. In that method call, you provide a block to be called by AVFoundation as it processes each frame of video, and the block’s sole parameter is a AVAsynchronousCIImageFilteringRequest object. Use that object both to the video frame image to be filtered and allows you to return a filtered image to AVFoundation for display or export. Listing 1 shows an example of applying a filter to an asset.
let filter = CIFilter(name: "CIGaussianBlur")!
let composition = AVVideoComposition(asset: asset, applyingCIFiltersWithHandler: { request in
// Clamp to avoid blurring transparent pixels at the image edges
let source = request.sourceImage.imageByClampingToExtent()
filter.setValue(source, forKey: kCIInputImageKey)
// Vary filter parameters based on video timing
let seconds = CMTimeGetSeconds(request.compositionTime)
filter.setValue(seconds * 10.0, forKey: kCIInputRadiusKey)
// Crop the blurred output to the bounds of the original image
let output = filter.outputImage!.imageByCroppingToRect(request.sourceImage.extent)
// Provide the filter output to the composition
request.finishWithImage(output, context: nil)
})
There is a tutorial in Objective C that may be a good resource as well.

Related

AVExportSession | Add UIView with GIFs and drawings on a video

Objective : I have a Video over which I have a UIView which contains animated GIFs(not locally stored, but using giphy api), Texts, or hand drawings. I want to export this along with the image in a single video.
What I did :
I created a UIView on which the animations are. Then converted that to CALayer and added to video with AVMutableVideoCompotion.
Problem : The UIView with animations is being converted to an Image instead of a video. How can I solve this.
Below is the Program for my export session. Any pointers will be really helpful.
func convertVideoAndSaveTophotoLibrary(videoURL: URL) {
let file = FileManager.shared.getDocumentDirectory(path: currentFilename)
FileManager.shared.clearPreviousFiles(withPath: file.path)
// File to composit
let asset = AVURLAsset(url: videoURL as URL)
let composition = AVMutableComposition.init()
composition.addMutableTrack(withMediaType: AVMediaType.video, preferredTrackID: kCMPersistentTrackID_Invalid)
let clipVideoTrack = asset.tracks(withMediaType: AVMediaType.video)[0]
// Rotate to potrait
let transformer = AVMutableVideoCompositionLayerInstruction(assetTrack: clipVideoTrack)
let videoTransform:CGAffineTransform = clipVideoTrack.preferredTransform
//fix orientation
var videoAssetOrientation_ = UIImage.Orientation.up
var isVideoAssetPortrait_ = false
if videoTransform.a == 0 && videoTransform.b == 1.0 && videoTransform.c == -1.0 && videoTransform.d == 0 {
videoAssetOrientation_ = UIImage.Orientation.right
isVideoAssetPortrait_ = true
}
if videoTransform.a == 0 && videoTransform.b == -1.0 && videoTransform.c == 1.0 && videoTransform.d == 0 {
videoAssetOrientation_ = UIImage.Orientation.left
isVideoAssetPortrait_ = true
}
if videoTransform.a == 1.0 && videoTransform.b == 0 && videoTransform.c == 0 && videoTransform.d == 1.0 {
videoAssetOrientation_ = UIImage.Orientation.up
}
if videoTransform.a == -1.0 && videoTransform.b == 0 && videoTransform.c == 0 && videoTransform.d == -1.0 {
videoAssetOrientation_ = UIImage.Orientation.down;
}
transformer.setTransform(clipVideoTrack.preferredTransform, at: CMTime.zero)
transformer.setOpacity(0.0, at: asset.duration)
//adjust the render size if neccessary
var naturalSize: CGSize
if(isVideoAssetPortrait_){
naturalSize = CGSize(width: clipVideoTrack.naturalSize.height, height: clipVideoTrack.naturalSize.width)
} else {
naturalSize = clipVideoTrack.naturalSize;
}
var renderWidth: CGFloat!
var renderHeight: CGFloat!
renderWidth = naturalSize.width
renderHeight = naturalSize.height
let parentlayer = CALayer()
let videoLayer = CALayer()
let watermarkLayer = CALayer()
let videoComposition = AVMutableVideoComposition()
videoComposition.renderSize = CGSize(width: renderWidth, height: renderHeight)
videoComposition.frameDuration = CMTimeMake(value: 1, timescale: 30)
videoComposition.renderScale = 1.0
//---------------------->>>>>> converting uiview to uiimage
watermarkLayer.contents = canvasView.asImage().cgImage
parentlayer.frame = CGRect(origin: CGPoint(x: 0, y: 0), size: naturalSize)
videoLayer.frame = CGRect(origin: CGPoint(x: 0, y: 0), size: naturalSize)
watermarkLayer.frame = CGRect(origin: CGPoint(x: 0, y: 0), size: naturalSize)
parentlayer.addSublayer(videoLayer)
parentlayer.addSublayer(watermarkLayer)
//---------------------->>>>>> Add view to video
videoComposition.animationTool = AVVideoCompositionCoreAnimationTool(postProcessingAsVideoLayers: [videoLayer], in: parentlayer)
let instruction = AVMutableVideoCompositionInstruction()
instruction.timeRange = CMTimeRangeMake(start: CMTime.zero, duration: CMTimeMakeWithSeconds(60, preferredTimescale: 30))
instruction.layerInstructions = [transformer]
videoComposition.instructions = [instruction]
let exporter = AVAssetExportSession.init(asset: asset, presetName: AVAssetExportPresetHighestQuality)
exporter?.outputFileType = AVFileType.mp4
exporter?.outputURL = file
exporter?.videoComposition = videoComposition
exporter?.shouldOptimizeForNetworkUse = true
exporter!.exportAsynchronously(completionHandler: {() -> Void in
if exporter?.status == .completed {
let outputURL: URL? = exporter?.outputURL
self.saveToPhotoLibrary(url: outputURL!)
}
})
}
Converting UIView to UIimage
extension UIView {
func asImage() -> UIImage {
let renderer = UIGraphicsImageRenderer(bounds: bounds)
return renderer.image { rendererContext in
layer.render(in: rendererContext.cgContext)
}
}
}
Code for Adding the GIF(I am using the Giphy API here), so the gif is downloaded and then added
func didSelectMedia(giphyViewController: GiphyViewController, media: GPHMedia) {
addMedia(media: media)
giphyViewController.dismiss(animated: true) { [weak self] in
self?.giphy = nil
}
}
// GPHMediaView is a subclass of UIImageView
func addMedia(media: GPHMedia) {
let mediaView = GPHMediaView()
mediaView.media = media
mediaView.contentMode = .scaleAspectFill
mediaView.frame.size = CGSize(width: 150, height: 150)
mediaView.center = canvasView.center
canvasView.addSubview(mediaView)
print(mediaView.frame)
self.addGesturesTo(mediaView)
}
What I am getting: The cat over the video is a gif. But sadly all i get is one frame. Now I know that is because I am converting the view to image. But that's the solution I need to know. How do I have the gif merged to the video.
You have two ways to archive this. First you can convert gif to video and add it to composition, but you lose alpha channel. Second way and more relevant is to add CAKeyframeAnimation on gif layer. To do this you should get all image frames from gif and put it all to key CAKeyframeAnimation.values and set duration which equal to framesCount * framesPerSecond.
class func makeContentAnimation(beginTime: Double, values: [Any], frameRate: Double) -> CAKeyframeAnimation {
let animation = CAKeyframeAnimation(keyPath: "contents")
animation.values = values
animation.beginTime = beginTime.isZero ? AVCoreAnimationBeginTimeAtZero : beginTime
animation.duration = frameRate * Double(values.count)
animation.isRemovedOnCompletion = false
animation.repeatCount = .infinity
return animation
}

AVExportSession exporting video super slow

I am trying to simply increase the speed of my exporting of my merged video.
Here is the code: //from my extensive research online and on SO, I have pretty much come down to the preset PassThrough makes it super fast, however as I wrote in a comment in the code, my merging code does not seem to work with that preset for export :/
static func videoCompositionInstruction(_ track: AVCompositionTrack, asset: AVAsset)
-> AVMutableVideoCompositionLayerInstruction {
let instruction = AVMutableVideoCompositionLayerInstruction(assetTrack: track)
let assetTrack = asset.tracks(withMediaType: .video)[0]
let transform = assetTrack.preferredTransform
let assetInfo = orientationFromTransform(transform)
var scaleToFitRatio = 1080 / assetTrack.naturalSize.width
if assetInfo.isPortrait {
scaleToFitRatio = 1080 / assetTrack.naturalSize.height
let scaleFactor = CGAffineTransform(scaleX: scaleToFitRatio, y: scaleToFitRatio)
var finalTransform = assetTrack.preferredTransform.concatenating(scaleFactor)
//was needed in my case (if video not taking entire screen and leaving some parts black - don't know when actually needed so you'll have to try and see when it's needed)
if assetInfo.orientation == .rightMirrored || assetInfo.orientation == .leftMirrored {
finalTransform = finalTransform.translatedBy(x: -transform.ty, y: 0)
}
instruction.setTransform(finalTransform, at: CMTime.zero)
} else {
let scaleFactor = CGAffineTransform(scaleX: scaleToFitRatio, y: scaleToFitRatio)
var concat = assetTrack.preferredTransform.concatenating(scaleFactor)
.concatenating(CGAffineTransform(translationX: 0, y: UIScreen.main.bounds.width / 2))
if assetInfo.orientation == .down {
let fixUpsideDown = CGAffineTransform(rotationAngle: CGFloat(Double.pi))
let windowBounds = UIScreen.main.bounds
let yFix = assetTrack.naturalSize.height + windowBounds.height
let centerFix = CGAffineTransform(translationX: assetTrack.naturalSize.width, y: yFix)
concat = fixUpsideDown.concatenating(centerFix).concatenating(scaleFactor)
}
instruction.setTransform(concat, at: CMTime.zero)
}
return instruction
}
static func orientationFromTransform(_ transform: CGAffineTransform)
-> (orientation: UIImage.Orientation, isPortrait: Bool) {
var assetOrientation = UIImage.Orientation.up
var isPortrait = false
if transform.a == 0 && transform.b == 1.0 && transform.c == -1.0 && transform.d == 0 {
assetOrientation = .right
isPortrait = true
} else if transform.a == 0 && transform.b == 1.0 && transform.c == 1.0 && transform.d == 0 {
assetOrientation = .rightMirrored
isPortrait = true
} else if transform.a == 0 && transform.b == -1.0 && transform.c == 1.0 && transform.d == 0 {
assetOrientation = .left
isPortrait = true
} else if transform.a == 0 && transform.b == -1.0 && transform.c == -1.0 && transform.d == 0 {
assetOrientation = .leftMirrored
isPortrait = true
} else if transform.a == 1.0 && transform.b == 0 && transform.c == 0 && transform.d == 1.0 {
assetOrientation = .up
} else if transform.a == -1.0 && transform.b == 0 && transform.c == 0 && transform.d == -1.0 {
assetOrientation = .down
}
return (assetOrientation, isPortrait)
}
func mergeVideosTestSQ(arrayVideos:[AVAsset], completion:#escaping (URL?, Error?) -> ()) {
let mixComposition = AVMutableComposition()
var instructions: [AVMutableVideoCompositionLayerInstruction] = []
var insertTime = CMTime(seconds: 0, preferredTimescale: 1)
/// for each URL add the video and audio tracks and their duration to the composition
for sourceAsset in arrayVideos {
let frameRange = CMTimeRange(start: CMTime(seconds: 0, preferredTimescale: 1), duration: sourceAsset.duration)
guard
let nthVideoTrack = mixComposition.addMutableTrack(withMediaType: .video, preferredTrackID: Int32(kCMPersistentTrackID_Invalid)),
let nthAudioTrack = mixComposition.addMutableTrack(withMediaType: .audio, preferredTrackID: Int32(kCMPersistentTrackID_Invalid)), //0 used to be kCMPersistentTrackID_Invalid
let assetVideoTrack = sourceAsset.tracks(withMediaType: .video).first
else {
print("didnt work")
return
}
var assetAudioTrack: AVAssetTrack?
assetAudioTrack = sourceAsset.tracks(withMediaType: .audio).first
print(assetAudioTrack, ",-- assetAudioTrack???", assetAudioTrack?.asset, "<-- hes", sourceAsset)
do {
try nthVideoTrack.insertTimeRange(frameRange, of: assetVideoTrack, at: insertTime)
try nthAudioTrack.insertTimeRange(frameRange, of: assetAudioTrack!, at: insertTime)
//instructions:
let nthInstruction = MainCamVC.videoCompositionInstruction(nthVideoTrack, asset: sourceAsset)
nthInstruction.setOpacity(0.0, at: CMTimeAdd(insertTime, sourceAsset.duration)) //sourceasset.duration
instructions.append(nthInstruction)
insertTime = insertTime + sourceAsset.duration //sourceAsset.duration
} catch {
DispatchQueue.main.async {
print("didnt wor2k")
}
}
}
let mainInstruction = AVMutableVideoCompositionInstruction()
mainInstruction.timeRange = CMTimeRange(start: CMTime(seconds: 0, preferredTimescale: 1), duration: insertTime)
mainInstruction.layerInstructions = instructions
let mainComposition = AVMutableVideoComposition()
mainComposition.instructions = [mainInstruction]
mainComposition.frameDuration = CMTimeMake(value: 1, timescale: 30)
mainComposition.renderSize = CGSize(width: 1080, height: 1920)
let outputFileURL = URL(fileURLWithPath: NSTemporaryDirectory() + "merge.mp4")
//below to clear the video form docuent folder for new vid...
let fileManager = FileManager()
try? fileManager.removeItem(at: outputFileURL)
/// try to start an export session and set the path and file type
if let exportSession = AVAssetExportSession(asset: mixComposition, presetName: AVAssetExportPresetHighestQuality) { //DOES NOT WORK WITH AVAssetExportPresetPassthrough
exportSession.outputFileType = .mov
exportSession.outputURL = outputFileURL
exportSession.videoComposition = mainComposition
exportSession.shouldOptimizeForNetworkUse = true
/// try to export the file and handle the status cases
exportSession.exportAsynchronously {
if let url = exportSession.outputURL{
completion(url, nil)
}
if let error = exportSession.error {
completion(nil, error)
}
}
}
}
Note I have instructions in order to preserve correct orientations.
Thanks for any help! I just need it to be faster, it roughly takes videoDuration/2 seconds to export in time to export...
After implementing your code into my project, it seems what is making your export slow, would be the way you handle the renderSize, as well as the resolution of the video. On top of that perhaps using a lower preset of quality may make it higher.
Specicially I would note this part:
let mainInstruction = AVMutableVideoCompositionInstruction()
mainInstruction.timeRange = CMTimeRange(start: CMTime(seconds: 0, preferredTimescale: 1), duration: insertTime)
mainInstruction.layerInstructions = instructions
let mainComposition = AVMutableVideoComposition()
mainComposition.instructions = [mainInstruction]
mainComposition.frameDuration = CMTimeMake(value: 1, timescale: 30)
mainComposition.renderSize = CGSize(width: 1080, height: 1920)
Changing the renderSize to = the videoCompositions.size (may be different name for your project) does the trick.
Then in the exporting place, I suggest changing this part:
/// try to start an export session and set the path and file type
if let exportSession = AVAssetExportSession(asset: mixComposition, presetName: AVAssetExportPresetHighestQuality) { //DOES NOT WORK WITH AVAssetExportPresetPassthrough
exportSession.outputFileType = .mov
exportSession.outputURL = outputFileURL
exportSession.videoComposition = mainComposition
exportSession.shouldOptimizeForNetworkUse = true
/// try to export the file and handle the status cases
exportSession.exportAsynchronously {
if let url = exportSession.outputURL{
completion(url, nil)
}
if let error = exportSession.error {
completion(nil, error)
}
}
}
As I said before, change it to presetQuality one lower or so. It will vastly improve the speed! Some transformations were also duplicated as well, something to look out for!

AVMutableComposition resizing issue

I'm trying to render an image into a video captured with the front camera using AVMutableComposition. The size of the resulting video (including the image) is perfectly fine.
However, the initial video will be resized as shown in this picture:
I'm using the NextLevelSessionExporter and this is my code snippet:
// * MARK - Creating composition
/// Create AVMutableComposition object. This object will hold the AVMutableCompositionTrack instances.
let mainMutableComposition = AVMutableComposition()
/// Creating an empty video track
let videoTrack = mainMutableComposition.addMutableTrack(withMediaType: AVMediaType.video, preferredTrackID: kCMPersistentTrackID_Invalid)
let videoAssetTrack = videoAsset.tracks(withMediaType: AVMediaType.video)[0]
do {
//Adding the video track
try videoTrack?.insertTimeRange(CMTimeRange(start: kCMTimeZero, duration: videoAsset.duration), of: videoAsset.tracks(withMediaType: AVMediaType.video).first!, at: kCMTimeZero)
} catch {
completion(false, nil)
}
/// Adding audio if user wants to.
if withAudio {
do {
//Adding the video track
let audio = videoAsset.tracks(withMediaType: AVMediaType.audio).first
if audio != nil {
let audioTrack = mainMutableComposition.addMutableTrack(withMediaType: AVMediaType.audio, preferredTrackID: kCMPersistentTrackID_Invalid)
try audioTrack?.insertTimeRange(CMTimeRange(start: kCMTimeZero, duration: videoAsset.duration), of: audio!, at: kCMTimeZero)
}
} catch {
completion(false, nil)
}
}
// * MARK - Composition is ready ----------
// Create AVMutableVideoCompositionInstruction
let compositionInstructions = AVMutableVideoCompositionInstruction()
compositionInstructions.timeRange = CMTimeRange(start: kCMTimeZero, duration: videoAsset.duration)
// Create an AvmutableVideoCompositionLayerInstruction
let videoLayerInstruction = AVMutableVideoCompositionLayerInstruction.init(assetTrack: videoTrack!)
videoLayerInstruction.setTransform(videoAssetTrack.preferredTransform, at: kCMTimeZero)
compositionInstructions.layerInstructions = [videoLayerInstruction]
//Add instructions
let videoComposition = AVMutableVideoComposition()
let naturalSize : CGSize = videoAssetTrack.naturalSize
///Rendering image into video
let renderWidth = naturalSize.width
let renderHeight = naturalSize.height
//Assigning instructions and rendering size
videoComposition.renderSize = CGSize(width: renderWidth, height: renderHeight)
videoComposition.instructions = [compositionInstructions]
videoComposition.frameDuration = CMTime(value: 1, timescale: Int32((videoTrack?.nominalFrameRate)!))
//Applying image to instruction
self.applyVideoImage(to: videoComposition, withSize: naturalSize, image: image)
// Getting the output path
let documentsURL = FileManager.default.urls(for: .documentDirectory, in: .userDomainMask).first
let outputPath = documentsURL?.appendingPathComponent("lastEditedVideo.mp4")
if FileManager.default.fileExists(atPath: (outputPath?.path)!) {
do {
try FileManager.default.removeItem(atPath: (outputPath?.path)!)
}
catch {
completion(false, nil)
}
}
// Create exporter
let exporter = NextLevelSessionExporter(withAsset: mainMutableComposition)
exporter.outputURL = outputPath
exporter.outputFileType = AVFileType.mp4
exporter.videoComposition = videoComposition
let compressionDict: [String: Any] = [
AVVideoAverageBitRateKey: NSNumber(integerLiteral: 2300000),
AVVideoProfileLevelKey: AVVideoProfileLevelH264BaselineAutoLevel as String
]
exporter.videoOutputConfiguration = [
AVVideoCodecKey: AVVideoCodecType.h264,
AVVideoWidthKey: NSNumber(integerLiteral: Int(naturalSize.width)),
AVVideoHeightKey: NSNumber(integerLiteral: Int(naturalSize.height)),
AVVideoCompressionPropertiesKey: compressionDict
]
exporter.audioOutputConfiguration = [
AVFormatIDKey: kAudioFormatMPEG4AAC,
AVEncoderBitRateKey: NSNumber(integerLiteral: 128000),
AVNumberOfChannelsKey: NSNumber(integerLiteral: 2),
AVSampleRateKey: NSNumber(value: Float(44100))
]
completion(true, exporter)
}
This is my applyVideoImage() function.
private func applyVideoImage(to composition: AVMutableVideoComposition, withSize size: CGSize, image: UIImage) { //Adds an image to a video composition
//Creating image layer
let overlayLayer = CALayer()
let overlayImage: UIImage = image
overlayLayer.contents = overlayImage.cgImage
overlayLayer.frame = CGRect(x: 0, y: 0, width: size.width, height: size.height)
overlayLayer.contentsGravity = kCAGravityResizeAspectFill
overlayLayer.masksToBounds = true
//Creating parent and video layer
let parentLayer = CALayer()
let videoLayer = CALayer()
parentLayer.frame = CGRect(x: 0, y: 0, width: size.width, height: size.height)
videoLayer.frame = CGRect(x: 0, y: 0, width: size.width, height: size.height)
parentLayer.addSublayer(videoLayer)
parentLayer.addSublayer(overlayLayer)
//Adding those layers to video
composition.animationTool = AVVideoCompositionCoreAnimationTool(postProcessingAsVideoLayer: videoLayer, in: parentLayer)
}
EDIT 1:
This bug only occurs when I'm exporting a mirrored video that has been captured with the front camera.
This is really tricky: You need to check the preferredTransform of the video track to determine wether it is a portrait video or not.
var videoAssetOrientation = UIImageOrientation.up
var isVideoAssetPortrait = false
var videoTransform = videoAssetTrack.preferredTransform
if needsMirroring == true {
isVideoAssetPortrait = true
}else if videoTransform.a == 0 && videoTransform.b == 1.0 && videoTransform.c == -1.0 && videoTransform.d == 0 {
videoAssetOrientation = .right
isVideoAssetPortrait = true
}else if videoTransform.a == 0 && videoTransform.b == -1.0 && videoTransform.c == 1.0 && videoTransform.d == 0 {
videoAssetOrientation = .left
isVideoAssetPortrait = true
}else if videoTransform.a == 1.0 && videoTransform.b == 0 && videoTransform.c == 0 && videoTransform.d == 1.0 {
videoAssetOrientation = .up
}else if videoTransform.a == -1.0 && videoTransform.b == 0 && videoTransform.c == 0 && videoTransform.d == -1.0 {
videoAssetOrientation = .down
}
//Add instructions
mainInstruction.layerInstructions = [videoLayerInstruction]
let mainCompositionInst = AVMutableVideoComposition()
let naturalSize : CGSize!
if isVideoAssetPortrait {
naturalSize = CGSize(width: videoAssetTrack.naturalSize.height, height: videoAssetTrack.naturalSize.width)
} else {
naturalSize = videoAssetTrack.naturalSize
}
Hope that helps.
Try applying a negative scale transform to flip the video when mirrored:
// Create an AvmutableVideoCompositionLayerInstruction
let videoLayerInstruction = AVMutableVideoCompositionLayerInstruction.init(assetTrack: videoTrack!)
let flipped = videoAssetTrack.preferredTransform.scaledBy(x: -1.0, y: 1.0)
videoLayerInstruction.setTransform(flipped, at: kCMTimeZero)
compositionInstructions.layerInstructions = [videoLayerInstruction]

Swift: Square video composition

I am following the below code for square video composition
func completeWithVideoAtURL(input: NSURL) {
let asset = AVAsset(url: input as URL)
let output = NSURL(fileURLWithPath: NSHomeDirectory() + "/Documents/Video.mp4")
let session = AVAssetExportSession(asset: asset, presetName: AVAssetExportPresetMediumQuality)!
session.videoComposition = self.squareVideoCompositionForAsset(asset: asset)
session.outputURL = output as URL
session.outputFileType = AVFileTypeMPEG4
session.shouldOptimizeForNetworkUse = true
session.exportAsynchronously(completionHandler: { () -> Void in
DispatchQueue.main.async(execute: { () -> Void in
// do something with the output
print("\(output)")
PHPhotoLibrary.shared().performChanges({
PHAssetChangeRequest.creationRequestForAssetFromVideo(atFileURL: output as URL)
}) { saved, error in
if saved {
print("saved to gallery")
}
}
})
})
}
func squareVideoCompositionForAsset(asset: AVAsset) -> AVVideoComposition {
let track = asset.tracks(withMediaType: AVMediaTypeVideo)[0]
let length = max(track.naturalSize.width, track.naturalSize.height)
var transform = track.preferredTransform
let size = track.naturalSize
let scale: CGFloat = (transform.a == -1 && transform.b == 0 && transform.c == 0 && transform.d == -1) ? -1 : 1 // check for inversion
transform = transform.translatedBy(x: scale * -(size.width - length) / 2, y: scale * -(size.height - length) / 2)
let transformer = AVMutableVideoCompositionLayerInstruction(assetTrack: track)
transformer.setTransform(transform, at: kCMTimeZero)
let instruction = AVMutableVideoCompositionInstruction()
instruction.timeRange = CMTimeRange(start: kCMTimeZero, duration: kCMTimePositiveInfinity)
instruction.layerInstructions = [transformer]
let composition = AVMutableVideoComposition()
composition.frameDuration = CMTime(value: 1, timescale: 30)
composition.renderSize = CGSize(width: length, height: length)
composition.instructions = [instruction]
return composition
}
From the squareVideoCompositionForAsset() function I take the max value for length between track.naturalSize.width & track.naturalSize.height cause I don't want to crop any partial part of the video. If I take min value, for portrait video it cropped the upper & lower portion of the video and for landscape video it cropped some left & right portion of the video.
For landscape video, output is okay
but for portrait video, output is like following image
the video gets left sided. Is it possible to center the video? Any assistance would be great and sorry for long explanation.
instead of this line
let scale: CGFloat = (transform.a == -1 && transform.b == 0 &&
transform.c == 0 && transform.d == -1) ? -1 : 1
I just used this
var scale = CGFloat()
if (transform.a == 0 && transform.b == 1 && transform.c == -1 && transform.d == 0) {
scale = -1
}
else if (transform.a == 0 && transform.b == -1 && transform.c == 1 && transform.d == 0) {
scale = -1
}
else if (transform.a == 1 && transform.b == 0 && transform.c == 0 && transform.d == 1) {
scale = 1
}
else if (transform.a == -1 && transform.b == 0 && transform.c == 0 && transform.d == -1) {
scale = 1
}
and it worked like a charm
Swift 4.2
func suqareCropVideo(videoURL: URL, withSide sideLength: CGFloat, completion: #escaping (_ resultURL: URL?, _ error: Error?) -> ()) {
let asset = AVAsset(url: videoURL)
if let assetVideoTrack = asset.tracks(withMediaType: .video).last {
let originalSize = assetVideoTrack.naturalSize
var scale: CGFloat
if originalSize.width < originalSize.height {
scale = sideLength / originalSize.width
} else {
scale = sideLength / originalSize.height
}
let scaledSize = CGSize(width: originalSize.width * scale, height: originalSize.height * scale)
let topLeft = CGPoint(x: sideLength * 0.5 - scaledSize.width * 0.5, y: sideLength * 0.5 - scaledSize.height * 0.5)
let layerInstruction = AVMutableVideoCompositionLayerInstruction(assetTrack: assetVideoTrack)
var orientationTransform = assetVideoTrack.preferredTransform
if (orientationTransform.tx == originalSize.width || orientationTransform.tx == originalSize.height) {
orientationTransform.tx = sideLength
}
if (orientationTransform.ty == originalSize.width || orientationTransform.ty == originalSize.height) {
orientationTransform.ty = sideLength
}
let transform = CGAffineTransform(scaleX: scale, y: scale).concatenating(CGAffineTransform(translationX: topLeft.x, y: topLeft.y)).concatenating(orientationTransform)
layerInstruction.setTransform(transform, at: .zero)
let instruction = AVMutableVideoCompositionInstruction()
instruction.layerInstructions = [layerInstruction]
instruction.timeRange = assetVideoTrack.timeRange
let videoComposition = AVMutableVideoComposition()
videoComposition.renderSize = CGSize(width: sideLength, height: sideLength)
videoComposition.renderScale = 1.0
videoComposition.frameDuration = CMTime(value: 1, timescale: 30)
videoComposition.instructions = [instruction]
if let export = AVAssetExportSession(asset: asset, presetName: AVAssetExportPresetHighestQuality) {
export.videoComposition = videoComposition
export.outputURL = NSURL.fileURL(withPath: "\(NSTemporaryDirectory())\(NSUUID().uuidString).mp4")
export.outputFileType = AVFileType.mp4
export.shouldOptimizeForNetworkUse = true
export.exportAsynchronously {
DispatchQueue.main.async {
if export.status == .completed {
completion(export.outputURL, nil)
} else {
completion(nil, export.error)
}
}
}
} else {
completion(nil, nil)
}
}
}

AVFoundation exporting orientation wrong

I'm attempting to combine an image and a video. I have them combining and exporting however it's rotated side ways.
Sorry for the bulk code paste. I've seen answers about applying a transform to compositionVideoTrack.preferredTransform however that does nothing. Adding to AVMutableVideoCompositionInstruction does nothing also.
I feel like this area is where things start to go wrong. here here:
// I feel like this loading here is the problem
let videoTrack = videoAsset.tracksWithMediaType(AVMediaTypeVideo)[0]
// because it makes our parentLayer and videoLayer sizes wrong
let videoSize = videoTrack.naturalSize
// this is returning 1920x1080, so it is rotating the video
print("\(videoSize.width) , \(videoSize.height)")
So by here our frame sizes are wrong for the rest of the method. Now when we try to go and create the overlay image layer the frame is not correct:
let aLayer = CALayer()
aLayer.contents = UIImage(named: "OverlayTestImageOverlay")?.CGImage
aLayer.frame = CGRectMake(0, 0, videoSize.width, videoSize.height)
aLayer.opacity = 1
Here is my complete method.
func combineImageVid() {
let path = NSBundle.mainBundle().pathForResource("SampleMovie", ofType:"MOV")
let fileURL = NSURL(fileURLWithPath: path!)
let videoAsset = AVURLAsset(URL: fileURL)
let mixComposition = AVMutableComposition()
let compositionVideoTrack = mixComposition.addMutableTrackWithMediaType(AVMediaTypeVideo, preferredTrackID: kCMPersistentTrackID_Invalid)
var clipVideoTrack = videoAsset.tracksWithMediaType(AVMediaTypeVideo)
do {
try compositionVideoTrack.insertTimeRange(CMTimeRangeMake(kCMTimeZero, videoAsset.duration), ofTrack: clipVideoTrack[0], atTime: kCMTimeZero)
}
catch _ {
print("failed to insertTimeRange")
}
compositionVideoTrack.preferredTransform = videoAsset.preferredTransform
// I feel like this loading here is the problem
let videoTrack = videoAsset.tracksWithMediaType(AVMediaTypeVideo)[0]
// because it makes our parentLayer and videoLayer sizes wrong
let videoSize = videoTrack.naturalSize
// this is returning 1920x1080, so it is rotating the video
print("\(videoSize.width) , \(videoSize.height)")
let aLayer = CALayer()
aLayer.contents = UIImage(named: "OverlayTestImageOverlay")?.CGImage
aLayer.frame = CGRectMake(0, 0, videoSize.width, videoSize.height)
aLayer.opacity = 1
let parentLayer = CALayer()
let videoLayer = CALayer()
parentLayer.frame = CGRectMake(0, 0, videoSize.width, videoSize.height)
videoLayer.frame = CGRectMake(0, 0, videoSize.width, videoSize.height)
parentLayer.addSublayer(videoLayer)
parentLayer.addSublayer(aLayer)
let videoComp = AVMutableVideoComposition()
videoComp.renderSize = videoSize
videoComp.frameDuration = CMTimeMake(1, 30)
videoComp.animationTool = AVVideoCompositionCoreAnimationTool(postProcessingAsVideoLayer: videoLayer, inLayer: parentLayer)
let instruction = AVMutableVideoCompositionInstruction()
instruction.timeRange = CMTimeRangeMake(kCMTimeZero, mixComposition.duration)
let mixVideoTrack = mixComposition.tracksWithMediaType(AVMediaTypeVideo)[0]
mixVideoTrack.preferredTransform = CGAffineTransformMakeRotation(CGFloat(M_PI * 90.0 / 180))
let layerInstruction = AVMutableVideoCompositionLayerInstruction(assetTrack: mixVideoTrack)
instruction.layerInstructions = [layerInstruction]
videoComp.instructions = [instruction]
// create new file to receive data
let dirPaths = NSSearchPathForDirectoriesInDomains(.DocumentDirectory, .UserDomainMask, true)
let docsDir: AnyObject = dirPaths[0]
let movieFilePath = docsDir.stringByAppendingPathComponent("result.mov")
let movieDestinationUrl = NSURL(fileURLWithPath: movieFilePath)
do {
try NSFileManager.defaultManager().removeItemAtPath(movieFilePath)
}
catch _ {}
// use AVAssetExportSession to export video
let assetExport = AVAssetExportSession(asset: mixComposition, presetName:AVAssetExportPresetHighestQuality)
assetExport?.videoComposition = videoComp
assetExport!.outputFileType = AVFileTypeQuickTimeMovie
assetExport!.outputURL = movieDestinationUrl
assetExport!.exportAsynchronouslyWithCompletionHandler({
switch assetExport!.status{
case AVAssetExportSessionStatus.Failed:
print("failed \(assetExport!.error)")
case AVAssetExportSessionStatus.Cancelled:
print("cancelled \(assetExport!.error)")
default:
print("Movie complete")
// play video
NSOperationQueue.mainQueue().addOperationWithBlock({ () -> Void in
print(movieDestinationUrl)
})
}
})
}
This is what I'm getting exported:
I tried adding these two methods in order to rotate the video:
class func videoCompositionInstructionForTrack(track: AVCompositionTrack, asset: AVAsset) -> AVMutableVideoCompositionLayerInstruction {
let instruction = AVMutableVideoCompositionLayerInstruction(assetTrack: track)
let assetTrack = asset.tracksWithMediaType(AVMediaTypeVideo)[0]
let transform = assetTrack.preferredTransform
let assetInfo = orientationFromTransform(transform)
var scaleToFitRatio = UIScreen.mainScreen().bounds.width / assetTrack.naturalSize.width
if assetInfo.isPortrait {
scaleToFitRatio = UIScreen.mainScreen().bounds.width / assetTrack.naturalSize.height
let scaleFactor = CGAffineTransformMakeScale(scaleToFitRatio, scaleToFitRatio)
instruction.setTransform(CGAffineTransformConcat(assetTrack.preferredTransform, scaleFactor),
atTime: kCMTimeZero)
} else {
let scaleFactor = CGAffineTransformMakeScale(scaleToFitRatio, scaleToFitRatio)
var concat = CGAffineTransformConcat(CGAffineTransformConcat(assetTrack.preferredTransform, scaleFactor), CGAffineTransformMakeTranslation(0, UIScreen.mainScreen().bounds.width / 2))
if assetInfo.orientation == .Down {
let fixUpsideDown = CGAffineTransformMakeRotation(CGFloat(M_PI))
let windowBounds = UIScreen.mainScreen().bounds
let yFix = assetTrack.naturalSize.height + windowBounds.height
let centerFix = CGAffineTransformMakeTranslation(assetTrack.naturalSize.width, yFix)
concat = CGAffineTransformConcat(CGAffineTransformConcat(fixUpsideDown, centerFix), scaleFactor)
}
instruction.setTransform(concat, atTime: kCMTimeZero)
}
return instruction
}
class func orientationFromTransform(transform: CGAffineTransform) -> (orientation: UIImageOrientation, isPortrait: Bool) {
var assetOrientation = UIImageOrientation.Up
var isPortrait = false
if transform.a == 0 && transform.b == 1.0 && transform.c == -1.0 && transform.d == 0 {
assetOrientation = .Right
isPortrait = true
} else if transform.a == 0 && transform.b == -1.0 && transform.c == 1.0 && transform.d == 0 {
assetOrientation = .Left
isPortrait = true
} else if transform.a == 1.0 && transform.b == 0 && transform.c == 0 && transform.d == 1.0 {
assetOrientation = .Up
} else if transform.a == -1.0 && transform.b == 0 && transform.c == 0 && transform.d == -1.0 {
assetOrientation = .Down
}
return (assetOrientation, isPortrait)
}
The updated my combineImageVid() method adding this in
let instruction = AVMutableVideoCompositionInstruction()
instruction.timeRange = CMTimeRangeMake(kCMTimeZero, mixComposition.duration)
let mixVideoTrack = mixComposition.tracksWithMediaType(AVMediaTypeVideo)[0]
//let layerInstruction = AVMutableVideoCompositionLayerInstruction(assetTrack: mixVideoTrack)
//layerInstruction.setTransform(videoAsset.preferredTransform, atTime: kCMTimeZero)
let layerInstruction = videoCompositionInstructionForTrack(compositionVideoTrack, asset: videoAsset)
Which gives me this output:
So I'm getting closer however I feel that because the track is originally being loaded the wrong way, I need to address the issue there. Also, I don't know why the huge black box is there now. I thought maybe it was due to my image layer taking the bounds of the loaded video asset here:
aLayer.frame = CGRectMake(0, 0, videoSize.width, videoSize.height)
However changing that to some small width/height doesn't make a difference. I then thought about adding a crop rec to get rid of the black square but that didn't work either :(
Following Allens suggestions of not using these two methods:
class func videoCompositionInstructionForTrack(track: AVCompositionTrack, asset: AVAsset) -> AVMutableVideoCompositionLayerInstruction
class func orientationFromTransform(transform: CGAffineTransform) -> (orientation: UIImageOrientation, isPortrait: Bool)
But updating my original method to look like this:
videoLayer.frame = CGRectMake(0, 0, videoSize.height, videoSize.width) //notice the switched width and height
...
videoComp.renderSize = CGSizeMake(videoSize.height,videoSize.width) //this make the final video in portrait
...
layerInstruction.setTransform(videoTrack.preferredTransform, atTime: kCMTimeZero) //important piece of information let composition know you want to rotate the original video in output
We are getting really close however the problem now seems to be editing renderSize. If I change it to anything other than the landscape size I get this:
here is the document for the orientation at Apple:
https://developer.apple.com/library/ios/qa/qa1744/_index.html
if your original video was taken in portrait mode iOS, it's nature size will still be landscape, but it comes with a metadata of rotate in the mov file. In order to rotate your video, you need to make changes to your 1st piece of code with the following:
videoLayer.frame = CGRectMake(0, 0, videoSize.height, videoSize.width) //notice the switched width and height
...
videoComp.renderSize = CGSizeMake(videoSize.height,videoSize.width) //this make the final video in portrait
...
layerInstruction.setTransform(videoTrack.preferredTransform, atTime: kCMTimeZero) //important piece of information let composition know you want to rotate the original video in output
Yes, you are really close!
Maybe U should check the videoTrack's preferredTransform so to give it a exact renderSize and transform:
CGAffineTransform transform = assetVideoTrack.preferredTransform;
CGFloat rotation = [self rotationWithTransform:transform];
//if been rotated
if (rotation != 0)
{
//if rotation is 360°
if (fabs((rotation - M_PI * 2)) >= valueOfError) {
CGFloat m = rotation / M_PI;
CGAffineTransform t1;
//rotation is 90° or 270°
if (fabs(m - 1/2.0) < valueOfError || fabs(m - 3/2.0) < valueOfError) {
self.mutableVideoComposition.renderSize = CGSizeMake(assetVideoTrack.naturalSize.height,assetVideoTrack.naturalSize.width);
t1 = CGAffineTransformMakeTranslation(assetVideoTrack.naturalSize.height, 0);
}
//rotation is 180°
if (fabs(m - 1.0) < valueOfError) {
t1 = CGAffineTransformMakeTranslation(assetVideoTrack.naturalSize.width, assetVideoTrack.naturalSize.height);
}
CGAffineTransform t2 = CGAffineTransformRotate(t1,rotation);
// CGAffineTransform transform = makeTransform(1.0, 1.0, 90, videoTrack.naturalSize.height, 0);
[passThroughLayer setTransform:t2 atTime:kCMTimeZero];
}
}
//convert transform to radian
- (CGFloat)rotationWithTransform:(CGAffineTransform)t
{
return atan2f(t.b, t.a);
}

Resources