AVAssetWriter在iOS上产生带有听不见的音频的视频，但在其他平台上不产生吗？

问题描述

我有一个AVAssetWriter管道设置，用于记录h264视频和AAC（LC）音频。音频源以10kHz的间隔（48个采样，保证）以48kHz的频率产生16位带符号整数MONO PCM音频。

AVAssetWriter音频接收器的转换格式设置为AAC 44.1kHz，64kbit / s MONO。

var channelLayout = AudioChannelLayout.init()
channelLayout.mChannelLayoutTag = kAudioChannelLayoutTag_Mono
let audioOutputSettings: [String: Any] = [
    AVFormatIDKey : kAudioFormatMPEG4AAC,AVSampleRateKey : 44100,AVNumberOfChannelsKey: 1,AVEncoderBitRateKey: 64000,AVChannelLayoutKey: NSData(bytes: &channelLayout,length: MemoryLayout.size(ofValue: channelLayout)),]

我自己在Objective-C ++中构造了音频CMSampleBuffer（音频源是WebRTC，C ++库）。请参见下面的代码段。

...

self.numberOfChannels = 1; // Use mono for the sake of brevity.
self.sampleRate = 48000;
self.numberOfSamples = 480; // Number of samples per 10ms
elapsedtimeMS = CACurrentMediaTime();

...

UInt32 sampleSize = sizeof(SInt16); // Data is 16-bit signed integer 
AudioStreamBasicDescription audioFormat;
audioFormat.mSampleRate = self.sampleRate;
audioFormat.mFormatID = kAudioFormatLinearPCM;
audioFormat.mFormatFlags = kAudioFormatFlagIsSignedInteger | kAudioFormatFlagsNativeEndian | kAudioFormatFlagIsPacked;
audioFormat.mFramesPerPacket = 1;
audioFormat.mChannelsPerFrame = self.numberOfChannels;
audioFormat.mBitsPerChannel = 8 * sampleSize;
audioFormat.mBytesPerFrame = audioFormat.mBytesPerPacket = audioFormat.mChannelsPerFrame * sampleSize;

CMFormatDescriptionRef sourceFormat = NULL;
status = CMAudioFormatDescriptionCreate(kcfAllocatorDefault,&audioFormat,sizeof(acl),&acl,NULL,&sourceFormat);
if(status != 0) {
    LOGEW << "Failed to create audio format description";
    return nil;
}

...

// Setup mixed audio buffer
UInt32 bufferByteSize = XMAudioCapturerBufferSize * self.numberOfChannels;
if(audioBuffer == nullptr) {
    audioBuffer = (uint8_t*)malloc(bufferByteSize);
    memset(audioBuffer,bufferByteSize);
}

// Try to store the (mixed) audio frame data (if any samples are available)
// Data is 16-bit signed integer PCM in 48000Hz
if (m_impl != nullptr && !m_impl->GetAudioFrame(audioBuffer)) {
    return NO;
}

audiobufferlist bufferList;
bufferList.mNumberBuffers = 1;
bufferList.mBuffers[0].mNumberChannels = self.numberOfChannels;
bufferList.mBuffers[0].mDataByteSize = bufferByteSize;
bufferList.mBuffers[0].mData = audioBuffer;

CMSampleTimingInfo timing = { CMTimeMake(1,self.sampleRate),CMTimeMake(self.elapsednumberOfSamples,kCMTimeInvalid };
//CMSampleTimingInfo timing = { CMTimeMake(1,CMTimeMake(elapsedtimeMS,1000),kCMTimeInvalid };

CMSampleBufferRef buffer;
status = CMSampleBufferCreate(kcfAllocatorDefault,false,sourceFormat,(CMItemCount)self.numberOfSamples * self.numberOfChannels,1,&timing,&buffer);
if(status != 0) {
    LOGEW << "Failed to allocate sample buffer";
    return NO;
}

status = CMSampleBufferSetDataBufferFromaudiobufferlist(buffer,kcfAllocatorDefault,&bufferList);
if(status != 0) {
    LOGEW << "Failed to convert audio buffer list into sample buffer";
    return NO;
}

...

CFRelease(buffer);
self.elapsednumberOfSamples += self.numberOfSamples * self.numberOfChannels;
elapsedtimeMS += XMAudioCapturerAudioFrameSizeMS; // ++10ms

CMSampleBuffer通过委托方法从与视频不同的线程传递到AVAssetWriter音频接收器。

为简洁起见，我省略了AVAssetWriter实现。值得一提的是，传递给AVAssetWriter的第一个样本是视频样本，并且会话以该视频样本的时间戳开始。所有音频样本都从时间“ 0”开始写入，并根据每10ms写入的样本量增加。

我附上了在iPhone上制作的视频。该视频无法在QuickTime或iOS上播放。但是它可以在Windows 10，macOS和Linux / Android上的所有经过测试的播放器上运行。

产生的示例文件（至少在达尔文没有音频）： https://drive.google.com/file/d/1pNHpu3fmgXCG2dmt4_yrVYr3oaR2B470/view?usp=sharing

有人知道为什么这甚至是个问题吗？我最初的想法是，达尔文平台上的AAC解码器有些挑剔，但话又说回来，我会在其他平台上看到类似的行为吗？

解决方法

暂无找到可以解决该程序问题的有效方法，小编努力寻找整理中！

如果你已经找到好的解决方法，欢迎将解决方案带上本链接一起发送给小编。

小编邮箱:dio#foxmail.com (将#修改为@）

audio avassetwriter ios ios