ThePain/Swiftgram
1
0
Fork 0
mirror of https://github.com/Swiftgram/Telegram-iOS.git synced 2025-08-02 00:17:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixed opus encoding completion

Browse Source
This commit is contained in:
Ilya Laktyushin 2019-08-12 05:22:13 +03:00
parent 41c468c139
commit b4541f7495
3 changed files with 201 additions and 212 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

232
submodules/TelegramUI/TelegramUI/Bridge Audio/TGBridgeAudioEncoder.m
View File

@ -81,13 +81,13 @@ typedef enum {
NSDictionary *outputSettings = @ NSDictionary *outputSettings = @
{ {
AVFormatIDKey: @(kAudioFormatLinearPCM), AVFormatIDKey: @(kAudioFormatLinearPCM),
AVSampleRateKey: @(TGBridgeAudioEncoderSampleRate), AVSampleRateKey: @(TGBridgeAudioEncoderSampleRate),
AVNumberOfChannelsKey: @1, AVNumberOfChannelsKey: @1,
AVLinearPCMBitDepthKey: @16, AVLinearPCMBitDepthKey: @16,
AVLinearPCMIsFloatKey: @false, AVLinearPCMIsFloatKey: @false,
AVLinearPCMIsBigEndianKey: @false, AVLinearPCMIsBigEndianKey: @false,
AVLinearPCMIsNonInterleaved: @false AVLinearPCMIsNonInterleaved: @false
}; };
_readerOutput = [AVAssetReaderAudioMixOutput assetReaderAudioMixOutputWithAudioTracks:asset.tracks audioSettings:outputSettings]; _readerOutput = [AVAssetReaderAudioMixOutput assetReaderAudioMixOutputWithAudioTracks:asset.tracks audioSettings:outputSettings];
@ -114,9 +114,9 @@ typedef enum {
static ATQueue *queue = nil; static ATQueue *queue = nil;
static dispatch_once_t onceToken; static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^ dispatch_once(&onceToken, ^
{ {
queue = [[ATQueue alloc] initWithName:@"org.telegram.opusAudioEncoderQueue"]; queue = [[ATQueue alloc] initWithName:@"org.telegram.opusAudioEncoderQueue"];
}); });
return queue; return queue;
} }
@ -126,74 +126,74 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
- (void)startWithCompletion:(void (^)(NSString *, int32_t))completion - (void)startWithCompletion:(void (^)(NSString *, int32_t))completion
{ {
[[TGBridgeAudioEncoder processingQueue] dispatch:^ [[TGBridgeAudioEncoder processingQueue] dispatch:^
{ {
_oggWriter = [[TGOggOpusWriter alloc] init]; _oggWriter = [[TGOggOpusWriter alloc] init];
if (![_oggWriter beginWithDataItem:_tempFileItem]) if (![_oggWriter beginWithDataItem:_tempFileItem])
{ {
[self cleanup]; [self cleanup];
return; return;
} }
[_assetReader startReading]; [_assetReader startReading];
while (_assetReader.status != AVAssetReaderStatusCompleted) while (_assetReader.status != AVAssetReaderStatusCompleted)
{ {
if (_assetReader.status == AVAssetReaderStatusReading) if (_assetReader.status == AVAssetReaderStatusReading)
{ {
CMSampleBufferRef nextBuffer = [_readerOutput copyNextSampleBuffer]; CMSampleBufferRef nextBuffer = [_readerOutput copyNextSampleBuffer];
if (nextBuffer) if (nextBuffer)
{ {
AudioBufferList abl; AudioBufferList abl;
CMBlockBufferRef blockBuffer; CMBlockBufferRef blockBuffer;
CMSampleBufferGetAudioBufferListWithRetainedBlockBuffer(nextBuffer, NULL, &abl, sizeof(abl), NULL, NULL, kCMSampleBufferFlag_AudioBufferList_Assure16ByteAlignment, &blockBuffer); CMSampleBufferGetAudioBufferListWithRetainedBlockBuffer(nextBuffer, NULL, &abl, sizeof(abl), NULL, NULL, kCMSampleBufferFlag_AudioBufferList_Assure16ByteAlignment, &blockBuffer);
[[TGBridgeAudioEncoder processingQueue] dispatch:^ [[TGBridgeAudioEncoder processingQueue] dispatch:^
{ {
[self _processBuffer:&abl.mBuffers[0]]; [self _processBuffer:&abl.mBuffers[0]];
CFRelease(nextBuffer); CFRelease(nextBuffer);
CFRelease(blockBuffer); CFRelease(blockBuffer);
}]; }];
} }
else else
{ {
[[TGBridgeAudioEncoder processingQueue] dispatch:^ [[TGBridgeAudioEncoder processingQueue] dispatch:^
{ {
if (_tailLength > 0) { if (_tailLength > 0) {
[_oggWriter writeFrame:(uint8_t *)_audioBuffer.bytes frameByteCount:(NSUInteger)_tailLength]; [_oggWriter writeFrame:(uint8_t *)_audioBuffer.bytes frameByteCount:(NSUInteger)_tailLength];
} }
}]; }];
break; break;
} }
} }
} }
[[TGBridgeAudioEncoder processingQueue] dispatch:^ [[TGBridgeAudioEncoder processingQueue] dispatch:^
{ {
TGFileDataItem *dataItemResult = nil; TGFileDataItem *dataItemResult = nil;
NSTimeInterval durationResult = 0.0; NSTimeInterval durationResult = 0.0;
NSUInteger totalBytes = 0; NSUInteger totalBytes = 0;
if (_assetReader.status == AVAssetReaderStatusCompleted) if (_assetReader.status == AVAssetReaderStatusCompleted)
{ {
NSLog(@"finished"); NSLog(@"finished");
if (_oggWriter != nil) if (_oggWriter != nil && [_oggWriter writeFrame:NULL frameByteCount:0])
{ {
dataItemResult = _tempFileItem; dataItemResult = _tempFileItem;
durationResult = [_oggWriter encodedDuration]; durationResult = [_oggWriter encodedDuration];
totalBytes = [_oggWriter encodedBytes]; totalBytes = [_oggWriter encodedBytes];
} }
[self cleanup]; [self cleanup];
} }
//TGLog(@"[TGBridgeAudioEncoder#%x convert time: %f ms]", self, (CFAbsoluteTimeGetCurrent() - startTime) * 1000.0); //TGLog(@"[TGBridgeAudioEncoder#%x convert time: %f ms]", self, (CFAbsoluteTimeGetCurrent() - startTime) * 1000.0);
if (completion != nil) if (completion != nil)
completion(dataItemResult.path, (int32_t)durationResult); completion(dataItemResult.path, (int32_t)durationResult);
}]; }];
}]; }];
} }
- (void)_processBuffer:(AudioBuffer const *)buffer - (void)_processBuffer:(AudioBuffer const *)buffer
@ -306,11 +306,11 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
[_queue dispatch:^ [_queue dispatch:^
{ {
_fileName = filePath; _fileName = filePath;
_length = [[[NSFileManager defaultManager] attributesOfItemAtPath:_fileName error:nil][NSFileSize] unsignedIntegerValue]; _length = [[[NSFileManager defaultManager] attributesOfItemAtPath:_fileName error:nil][NSFileSize] unsignedIntegerValue];
_fileExists = [[NSFileManager defaultManager] fileExistsAtPath:_fileName]; _fileExists = [[NSFileManager defaultManager] fileExistsAtPath:_fileName];
}]; }];
} }
return self; return self;
} }
@ -322,38 +322,38 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
- (void)moveToPath:(NSString *)path - (void)moveToPath:(NSString *)path
{ {
[_queue dispatch:^ [_queue dispatch:^
{ {
[[NSFileManager defaultManager] moveItemAtPath:_fileName toPath:path error:nil]; [[NSFileManager defaultManager] moveItemAtPath:_fileName toPath:path error:nil];
_fileName = path; _fileName = path;
}]; }];
} }
- (void)remove - (void)remove
{ {
[_queue dispatch:^ [_queue dispatch:^
{ {
[[NSFileManager defaultManager] removeItemAtPath:_fileName error:nil]; [[NSFileManager defaultManager] removeItemAtPath:_fileName error:nil];
}]; }];
} }
- (void)appendData:(NSData *)data - (void)appendData:(NSData *)data
{ {
[_queue dispatch:^ [_queue dispatch:^
{ {
if (!_fileExists) if (!_fileExists)
{ {
[[NSFileManager defaultManager] createFileAtPath:_fileName contents:nil attributes:nil]; [[NSFileManager defaultManager] createFileAtPath:_fileName contents:nil attributes:nil];
_fileExists = true; _fileExists = true;
} }
NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName]; NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName];
[file seekToEndOfFile]; [file seekToEndOfFile];
[file writeData:data]; [file writeData:data];
[file synchronizeFile]; [file synchronizeFile];
[file closeFile]; [file closeFile];
_length += data.length; _length += data.length;
[_data appendData:data]; [_data appendData:data];
}]; }];
} }
- (NSData *)readDataAtOffset:(NSUInteger)offset length:(NSUInteger)length - (NSData *)readDataAtOffset:(NSUInteger)offset length:(NSUInteger)length
@ -361,14 +361,14 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
__block NSData *data = nil; __block NSData *data = nil;
[_queue dispatch:^ [_queue dispatch:^
{ {
NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName]; NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName];
[file seekToFileOffset:(unsigned long long)offset]; [file seekToFileOffset:(unsigned long long)offset];
data = [file readDataOfLength:length]; data = [file readDataOfLength:length];
if (data.length != length) if (data.length != length)
//TGLog(@"Read data length mismatch"); //TGLog(@"Read data length mismatch");
[file closeFile]; [file closeFile];
} synchronous:true]; } synchronous:true];
return data; return data;
} }
@ -377,9 +377,9 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
{ {
__block NSUInteger result = 0; __block NSUInteger result = 0;
[_queue dispatch:^ [_queue dispatch:^
{ {
result = _length; result = _length;
} synchronous:true]; } synchronous:true];
return result; return result;
} }
@ -420,11 +420,11 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
static ATQueue *queue = nil; static ATQueue *queue = nil;
static dispatch_once_t onceToken; static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^ dispatch_once(&onceToken, ^
{ {
queue = [[ATQueue alloc] init]; queue = [[ATQueue alloc] init];
queue->_nativeQueue = dispatch_get_main_queue(); queue->_nativeQueue = dispatch_get_main_queue();
queue->_isMainQueue = true; queue->_isMainQueue = true;
}); });
return queue; return queue;
} }
@ -434,9 +434,9 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
static ATQueue *queue = nil; static ATQueue *queue = nil;
static dispatch_once_t onceToken; static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^ dispatch_once(&onceToken, ^
{ {
queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0)]; queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0)];
}); });
return queue; return queue;
} }
@ -446,9 +446,9 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
static ATQueue *queue = nil; static ATQueue *queue = nil;
static dispatch_once_t onceToken; static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^ dispatch_once(&onceToken, ^
{ {
queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0)]; queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0)];
}); });
return queue; return queue;
} }

106
submodules/TelegramUI/TelegramUI/ManagedAudioRecorder.swift
View File

@ -319,7 +319,7 @@ final class ManagedAudioRecorderContext {
self.idleTimerExtensionDisposable = (Signal<Void, NoError> { subscriber in self.idleTimerExtensionDisposable = (Signal<Void, NoError> { subscriber in
return pushIdleTimerExtension() return pushIdleTimerExtension()
} |> delay(5.0, queue: queue)).start() } |> delay(5.0, queue: queue)).start()
} }
deinit { deinit {
@ -406,19 +406,19 @@ final class ManagedAudioRecorderContext {
strongSelf.audioSessionAcquired(headset: state.isHeadsetConnected) strongSelf.audioSessionAcquired(headset: state.isHeadsetConnected)
} }
} }
}, deactivate: { [weak self] in }, deactivate: { [weak self] in
return Signal { subscriber in return Signal { subscriber in
queue.async { queue.async {
if let strongSelf = self { if let strongSelf = self {
strongSelf.hasAudioSession = false strongSelf.hasAudioSession = false
strongSelf.stop() strongSelf.stop()
strongSelf.recordingState.set(.stopped) strongSelf.recordingState.set(.stopped)
subscriber.putCompletion() subscriber.putCompletion()
}
} }
return EmptyDisposable
} }
return EmptyDisposable
}
}) })
} }
} }
@ -592,53 +592,57 @@ final class ManagedAudioRecorderContext {
} }
func takeData() -> RecordedAudioData? { func takeData() -> RecordedAudioData? {
var scaledSamplesMemory = malloc(100 * 2)! if self.oggWriter.writeFrame(nil, frameByteCount: 0) {
var scaledSamples: UnsafeMutablePointer<Int16> = scaledSamplesMemory.assumingMemoryBound(to: Int16.self) var scaledSamplesMemory = malloc(100 * 2)!
defer { var scaledSamples: UnsafeMutablePointer<Int16> = scaledSamplesMemory.assumingMemoryBound(to: Int16.self)
free(scaledSamplesMemory) defer {
} free(scaledSamplesMemory)
memset(scaledSamples, 0, 100 * 2); }
var waveform: Data? memset(scaledSamples, 0, 100 * 2);
var waveform: Data?
let count = self.compressedWaveformSamples.count / 2 let count = self.compressedWaveformSamples.count / 2
self.compressedWaveformSamples.withUnsafeMutableBytes { (samples: UnsafeMutablePointer<Int16>) -> Void in self.compressedWaveformSamples.withUnsafeMutableBytes { (samples: UnsafeMutablePointer<Int16>) -> Void in
for i in 0 ..< count { for i in 0 ..< count {
let sample = samples[i] let sample = samples[i]
let index = i * 100 / count let index = i * 100 / count
if (scaledSamples[index] < sample) { if (scaledSamples[index] < sample) {
scaledSamples[index] = sample; scaledSamples[index] = sample;
}
} }
}
var peak: Int16 = 0 var peak: Int16 = 0
var sumSamples: Int64 = 0 var sumSamples: Int64 = 0
for i in 0 ..< 100 { for i in 0 ..< 100 {
let sample = scaledSamples[i] let sample = scaledSamples[i]
if peak < sample { if peak < sample {
peak = sample peak = sample
}
sumSamples += Int64(sample)
} }
sumSamples += Int64(sample) var calculatedPeak: UInt16 = 0
} calculatedPeak = UInt16((Double(sumSamples) * 1.8 / 100.0))
var calculatedPeak: UInt16 = 0
calculatedPeak = UInt16((Double(sumSamples) * 1.8 / 100.0))
if calculatedPeak < 2500 { if calculatedPeak < 2500 {
calculatedPeak = 2500 calculatedPeak = 2500
}
for i in 0 ..< 100 {
let sample: UInt16 = UInt16(Int64(scaledSamples[i]))
let minPeak = min(Int64(sample), Int64(calculatedPeak))
let resultPeak = minPeak * 31 / Int64(calculatedPeak)
scaledSamples[i] = Int16(clamping: min(31, resultPeak))
}
let resultWaveform = AudioWaveform(samples: Data(bytes: scaledSamplesMemory, count: 100 * 2), peak: 31)
let bitstream = resultWaveform.makeBitstream()
waveform = AudioWaveform(bitstream: bitstream, bitsPerSample: 5).makeBitstream()
} }
for i in 0 ..< 100 { return RecordedAudioData(compressedData: self.dataItem.data(), duration: self.oggWriter.encodedDuration(), waveform: waveform)
let sample: UInt16 = UInt16(Int64(scaledSamples[i])) } else {
let minPeak = min(Int64(sample), Int64(calculatedPeak)) return nil
let resultPeak = minPeak * 31 / Int64(calculatedPeak)
scaledSamples[i] = Int16(clamping: min(31, resultPeak))
}
let resultWaveform = AudioWaveform(samples: Data(bytes: scaledSamplesMemory, count: 100 * 2), peak: 31)
let bitstream = resultWaveform.makeBitstream()
waveform = AudioWaveform(bitstream: bitstream, bitsPerSample: 5).makeBitstream()
} }
return RecordedAudioData(compressedData: self.dataItem.data(), duration: self.oggWriter.encodedDuration(), waveform: waveform)
} }
} }

47
submodules/TelegramUI/third-party/opusenc/opusenc.m vendored
View File

@ -426,39 +426,24 @@ static inline int writeOggPage(ogg_page *page, TGDataItem *fileItem)
pages_out++; pages_out++;
} }
op.packet = (unsigned char *)_packet; if (framePcmBytes != NULL) {
op.bytes = nbBytes; op.packet = (unsigned char *)_packet;
op.b_o_s = 0; op.bytes = nbBytes;
op.granulepos = enc_granulepos; op.b_o_s = 0;
if (op.e_o_s) op.granulepos = enc_granulepos;
{
/* We compute the final GP as ceil(len*48k/input_rate). When a resampling
decoder does the matching floor(len*input/48k) conversion the length will
be exactly the same as the input.
*/
op.granulepos = ((total_samples * 48000 + rate - 1) / rate) + header.preskip;
}
op.packetno = 2 + _packetId;
ogg_stream_packetin(&os, &op);
last_segments += size_segments;
/* The downside of early reading is if the input is an exact if (op.e_o_s)
multiple of the frame_size you'll get an extra frame that needs {
to get cropped off. The downside of late reading is added delay. /* We compute the final GP as ceil(len*48k/input_rate). When a resampling
If your ogg_delay is 120ms or less we'll assume you want the decoder does the matching floor(len*input/48k) conversion the length will
low delay behavior. be exactly the same as the input.
*/ */
/*if ((!op.e_o_s) && max_ogg_delay > 5760) op.granulepos = ((total_samples * 48000 + rate - 1) / rate) + header.preskip;
{ }
nb_samples = inopt.read_samples(inopt.readdata, input, frame_size); op.packetno = 2 + _packetId;
total_samples += nb_samples; ogg_stream_packetin(&os, &op);
if (nb_samples < frame_size) last_segments += size_segments;
eos = 1;
if (nb_samples == 0)
op.e_o_s = 1;
} }
else
nb_samples = -1;*/
// If the stream is over or we're sure that the delayed flush will fire, go ahead and flush now to avoid adding delay // If the stream is over or we're sure that the delayed flush will fire, go ahead and flush now to avoid adding delay
while ((op.e_o_s || (enc_granulepos + (frame_size * 48000 / coding_rate) - last_granulepos > max_ogg_delay) || while ((op.e_o_s || (enc_granulepos + (frame_size * 48000 / coding_rate) - last_granulepos > max_ogg_delay) ||