前言
GPUImage详细解析在GPUImage文集,在前文GPUImage详细解析(八)视频合并混音中使用了github上的开源方式实现,这篇使用GPUImage原生的GPUImageMovieComposition
来合并视频。
核心思路
先加载视频信息,再配置轨道信息、视频操作指令和音频指令参数,创建GPUImageMovieComposition
类,设置输出目标为GPUImageMovieWriter
并开始处理,最后把处理完毕的数据写入手机。
效果展示
具体细节
1、加载视频信息
通过GCD的dispatch_group_notify
方法实现异步等待,加载完毕后调用-synchronizeWithEditor
方法。
- (void)loadAsset:(AVAsset *)asset withKeys:(NSArray *)assetKeysToLoad usingDispatchGroup:(dispatch_group_t)dispatchGroup
{
dispatch_group_enter(dispatchGroup);
[asset loadValuesAsynchronouslyForKeys:assetKeysToLoad completionHandler:^(){
// 测试是否成功加载
BOOL bSuccess = YES;
for (NSString *key in assetKeysToLoad) {
NSError *error;
if ([asset statusOfValueForKey:key error:&error] == AVKeyValueStatusFailed) {
NSLog(@"Key value loading failed for key:%@ with error: %@", key, error);
bSuccess = NO;
break;
}
}
if (![asset isComposable]) {
NSLog(@"Asset is not composable");
bSuccess = NO;
}
if (bSuccess && CMTimeGetSeconds(asset.duration) > 5) {
[self.clips addObject:asset];
[self.clipTimeRanges addObject:[NSValue valueWithCMTimeRange:CMTimeRangeMake(CMTimeMakeWithSeconds(0, 1), CMTimeMakeWithSeconds(5, 1))]];
}
else {
NSLog(@"error ");
}
dispatch_group_leave(dispatchGroup);
}];
}
2、配置轨道信息、音视频指令
- AVMutableVideoCompositionInstruction 视频操作指令
- AVMutableVideoCompositionLayerInstruction视频轨道操作指令,需要添加到AVMutableVideoCompositionInstruction
- AVMutableAudioMixInputParameters音频操作参数
- AVMutableComposition 包含多个轨道的媒体信息,可以添加、删除轨道
- AVMutableVideoComposition视频操作指令集合
- (void)buildCompositionObjectsForPlayback
{
if ( (self.clips == nil) || [self.clips count] == 0 ) {
self.composition = nil;
self.videoComposition = nil;
return;
}
CGSize videoSize = [[self.clips objectAtIndex:0] naturalSize];
AVMutableComposition *composition = [AVMutableComposition composition];
AVMutableVideoComposition *videoComposition = nil;
AVMutableAudioMix *audioMix = nil;
composition.naturalSize = videoSize;
videoComposition = [AVMutableVideoComposition videoComposition];
audioMix = [AVMutableAudioMix audioMix];
[self buildTransitionComposition:composition andVideoComposition:videoComposition andAudioMix:audioMix];
if (videoComposition) {
// 通用属性
videoComposition.frameDuration = CMTimeMake(1, 30); // 30 fps
videoComposition.renderSize = videoSize;
}
self.composition = composition;
self.videoComposition = videoComposition;
self.audioMix = audioMix;
}
3、创建GPUImageMovieComposition类,设置GPUImage相关信息
GPUImageMovieComposition的创建依赖第2步生成的轨道信息、音视频指令。
设置音视频的输出target为GPUImageMovieWriter,并在GPUImageMovieWriter的completionBlock中把处理完毕的视频存入手机。
代码语言:javascript复制- (void)synchronizePlayerWithEditor
{
NSString *pathToMovie = [NSHomeDirectory() stringByAppendingPathComponent:@"Documents/Movie.m4v"];
unlink([pathToMovie UTF8String]);
NSURL *movieURL = [NSURL fileURLWithPath:pathToMovie];
self.movieWriter = [[GPUImageMovieWriter alloc] initWithMovieURL:movieURL size:CGSizeMake(640, 480)];
self.imageMovieComposition = [[GPUImageMovieComposition alloc] initWithComposition:self.editor.composition andVideoComposition:self.editor.videoComposition andAudioMix:self.editor.audioMix];
self.imageMovieComposition.playAtActualSpeed = YES;
self.imageMovieComposition.runBenchmark = YES;
[self.imageMovieComposition addTarget:self.movieWriter];
[self.imageMovieComposition addTarget:filter];
[self.imageMovieComposition enableSynchronizedEncodingUsingMovieWriter:self.movieWriter];
self.imageMovieComposition.audioEncodingTarget = self.movieWriter;
[_movieWriter startRecording];
[self.imageMovieComposition startProcessing];
CADisplayLink* dlink = [CADisplayLink displayLinkWithTarget:self selector:@selector(updateProgress)];
[dlink addToRunLoop:[NSRunLoop currentRunLoop] forMode:NSRunLoopCommonModes];
[dlink setPaused:NO];
__weak typeof(self) weakSelf = self;
[_movieWriter setCompletionBlock:^{
__strong typeof(self) strongSelf = weakSelf;
[strongSelf->_movieWriter finishRecording];
[strongSelf->_imageMovieComposition endProcessing];
ALAssetsLibrary *library = [[ALAssetsLibrary alloc] init];
if (UIVideoAtPathIsCompatibleWithSavedPhotosAlbum(pathToMovie))
{
[library writeVideoAtPathToSavedPhotosAlbum:movieURL completionBlock:^(NSURL *assetURL, NSError *error)
{
dispatch_async(dispatch_get_main_queue(), ^{
if (error) {
UIAlertView *alert = [[UIAlertView alloc] initWithTitle:@"视频保存失败" message:nil
delegate:nil cancelButtonTitle:@"OK" otherButtonTitles:nil];
[alert show];
} else {
UIAlertView *alert = [[UIAlertView alloc] initWithTitle:@"视频保存成功" message:nil
delegate:nil cancelButtonTitle:@"OK" otherButtonTitles:nil];
[alert show];
}
});
}];
}
else {
NSLog(@"error mssg)");
}
}];
}
总结
距离上一篇GPUImage文章更新GPUImage详细解析(九)图像的输入输出和滤镜通道已经有两个月,这一篇更多是了却当时未能的想法。
GPUImage进阶的各种滤镜使用是GPUImage的核心魅力所在,目前仅是在直播中用到GPUImage的美颜功能。
在研读GPUImage源码的过程中对OpenGL ES有了更深入的学习,也发现了GPUImage的部分问题,比如说在推流过程中添加滤镜导致推流视频闪烁的现象(这个问题来自于一个简书的网友,当时的图如下)。
大致结构图
demo在代码地址。
扩展
通过debug,单步执行代码得到以下结果 1、Camera申请buffer_d90; 2、filter1申请buffer_420; 3、fitler1渲染完毕,归还buffer_d90; 4、推流读取buffer_420的数据; 5、ImageView读取buffer_420的数据,渲染到屏幕(预览),归还buffer_420; 1、Camera申请buffer_420; 2、filter1申请buffer_d90; 3、fitler1渲染完毕,归还buffer_420; 4、推流读取buffer_d90的数据; 5、ImageView读取buffer_d90的数据,渲染到屏幕(预览),归还buffer_d90; Camera申请到了不同的buffer,GPU处理又是异步操作,这样推流读取像素数据的时候,可能会读取到两个buffer的数据。
解决方案1:
添加一个空白的filter,debug结果如下:
1、Camera申请buffer_5e0;
2、filter1申请buffer_ba0;
3、fitler1渲染完毕,归还buffer_5e0;
4、filter2申请buffer_5e0;
5、filter2渲染完毕,归还buffer_ba0;
6、推流读取buffer_5e0的数据;
7、ImageView读取buffer_5e0的数据,渲染到屏幕(预览),归还buffer_5e0;
1、Camera申请buffer_5e0;
2、filter1申请buffer_ba0;
3、fitler1渲染完毕,归还buffer_5e0;
4、filter2申请buffer_5e0;
5、filter2渲染完毕,归还buffer_ba0;
6、推流读取buffer_5e0的数据;
7、ImageView读取buffer_5e0的数据,渲染到屏幕(预览),归还buffer_5e0;
Camera两次申请的buffer是同一块,这样推流读取的像素数据来自同一个buffer。
解决方案2:
在GPUImageBuffer类添加一行代码
代码语言:javascript复制 glFinish();