bool EffectSBSMS::Process()
{
bool bGoodResult = true;
//Iterate over each track
//Track::All is needed because this effect needs to introduce silence in the group tracks to keep sync
this->CopyInputTracks(Track::All); // Set up mOutputTracks.
TrackListIterator iter(mOutputTracks);
Track* t;
mCurTrackNum = 0;
double maxDuration = 0.0;
// Must sync if selection length will change
bool mustSync = (rateStart != rateEnd);
Slide rateSlide(rateSlideType,rateStart,rateEnd);
Slide pitchSlide(pitchSlideType,pitchStart,pitchEnd);
mTotalStretch = rateSlide.getTotalStretch();
t = iter.First();
while (t != NULL) {
if (t->GetKind() == Track::Label &&
(t->GetSelected() || (mustSync && t->IsSyncLockSelected())) )
{
if (!ProcessLabelTrack(t)) {
bGoodResult = false;
break;
}
}
else if (t->GetKind() == Track::Wave && t->GetSelected() )
{
WaveTrack* leftTrack = (WaveTrack*)t;
//Get start and end times from track
mCurT0 = leftTrack->GetStartTime();
mCurT1 = leftTrack->GetEndTime();
//Set the current bounds to whichever left marker is
//greater and whichever right marker is less
mCurT0 = wxMax(mT0, mCurT0);
mCurT1 = wxMin(mT1, mCurT1);
// Process only if the right marker is to the right of the left marker
if (mCurT1 > mCurT0) {
sampleCount start;
sampleCount end;
start = leftTrack->TimeToLongSamples(mCurT0);
end = leftTrack->TimeToLongSamples(mCurT1);
WaveTrack* rightTrack = NULL;
if (leftTrack->GetLinked()) {
double t;
rightTrack = (WaveTrack*)(iter.Next());
//Adjust bounds by the right tracks markers
t = rightTrack->GetStartTime();
t = wxMax(mT0, t);
mCurT0 = wxMin(mCurT0, t);
t = rightTrack->GetEndTime();
t = wxMin(mT1, t);
mCurT1 = wxMax(mCurT1, t);
//Transform the marker timepoints to samples
start = leftTrack->TimeToLongSamples(mCurT0);
end = leftTrack->TimeToLongSamples(mCurT1);
mCurTrackNum++; // Increment for rightTrack, too.
}
sampleCount trackStart = leftTrack->TimeToLongSamples(leftTrack->GetStartTime());
sampleCount trackEnd = leftTrack->TimeToLongSamples(leftTrack->GetEndTime());
// SBSMS has a fixed sample rate - we just convert to its sample rate and then convert back
float srTrack = leftTrack->GetRate();
float srProcess = bLinkRatePitch?srTrack:44100.0;
// the resampler needs a callback to supply its samples
ResampleBuf rb;
sampleCount maxBlockSize = leftTrack->GetMaxBlockSize();
rb.blockSize = maxBlockSize;
rb.buf = (audio*)calloc(rb.blockSize,sizeof(audio));
rb.leftTrack = leftTrack;
rb.rightTrack = rightTrack?rightTrack:leftTrack;
rb.leftBuffer = (float*)calloc(maxBlockSize,sizeof(float));
rb.rightBuffer = (float*)calloc(maxBlockSize,sizeof(float));
// Samples in selection
sampleCount samplesIn = end-start;
// Samples for SBSMS to process after resampling
sampleCount samplesToProcess = (sampleCount) ((float)samplesIn*(srProcess/srTrack));
SlideType outSlideType;
SBSMSResampleCB outResampleCB;
sampleCount processPresamples = 0;
sampleCount trackPresamples = 0;
if(bLinkRatePitch) {
rb.bPitch = true;
outSlideType = rateSlideType;
outResampleCB = resampleCB;
rb.offset = start;
rb.end = end;
rb.iface = new SBSMSInterfaceSliding(&rateSlide,&pitchSlide,
bPitchReferenceInput,
samplesToProcess,0,
NULL);
} else {
rb.bPitch = false;
outSlideType = (srProcess==srTrack?SlideIdentity:SlideConstant);
outResampleCB = postResampleCB;
rb.ratio = srProcess/srTrack;
rb.quality = new SBSMSQuality(&SBSMSQualityStandard);
rb.resampler = new Resampler(resampleCB, &rb, srProcess==srTrack?SlideIdentity:SlideConstant);
rb.sbsms = new SBSMS(rightTrack?2:1,rb.quality,true);
rb.SBSMSBlockSize = rb.sbsms->getInputFrameSize();
rb.SBSMSBuf = (audio*)calloc(rb.SBSMSBlockSize,sizeof(audio));
processPresamples = wxMin(rb.quality->getMaxPresamples(),
(long)((float)(start-trackStart)*(srProcess/srTrack)));
trackPresamples = wxMin(start-trackStart,
(long)((float)(processPresamples)*(srTrack/srProcess)));
rb.offset = start - trackPresamples;
rb.end = trackEnd;
rb.iface = new SBSMSEffectInterface(rb.resampler,
&rateSlide,&pitchSlide,
bPitchReferenceInput,
samplesToProcess,processPresamples,
rb.quality);
}
Resampler resampler(outResampleCB,&rb,outSlideType);
audio outBuf[SBSMSOutBlockSize];
float outBufLeft[2*SBSMSOutBlockSize];
float outBufRight[2*SBSMSOutBlockSize];
// Samples in output after SBSMS
sampleCount samplesToOutput = rb.iface->getSamplesToOutput();
// Samples in output after resampling back
sampleCount samplesOut = (sampleCount) ((float)samplesToOutput * (srTrack/srProcess));
// Duration in track time
double duration = (mCurT1-mCurT0) * mTotalStretch;
if(duration > maxDuration)
maxDuration = duration;
TimeWarper *warper = createTimeWarper(mCurT0,mCurT1,maxDuration,rateStart,rateEnd,rateSlideType);
SetTimeWarper(warper);
rb.outputLeftTrack = mFactory->NewWaveTrack(leftTrack->GetSampleFormat(),
leftTrack->GetRate());
if(rightTrack)
rb.outputRightTrack = mFactory->NewWaveTrack(rightTrack->GetSampleFormat(),
rightTrack->GetRate());
long pos = 0;
long outputCount = -1;
// process
while(pos<samplesOut && outputCount) {
long frames;
if(pos+SBSMSOutBlockSize>samplesOut) {
frames = samplesOut - pos;
} else {
frames = SBSMSOutBlockSize;
}
outputCount = resampler.read(outBuf,frames);
for(int i = 0; i < outputCount; i++) {
outBufLeft[i] = outBuf[i][0];
if(rightTrack)
outBufRight[i] = outBuf[i][1];
}
pos += outputCount;
rb.outputLeftTrack->Append((samplePtr)outBufLeft, floatSample, outputCount);
if(rightTrack)
rb.outputRightTrack->Append((samplePtr)outBufRight, floatSample, outputCount);
double frac = (double)pos/(double)samplesOut;
int nWhichTrack = mCurTrackNum;
if(rightTrack) {
nWhichTrack = 2*(mCurTrackNum/2);
if (frac < 0.5)
frac *= 2.0; // Show twice as far for each track, because we're doing 2 at once.
else {
nWhichTrack++;
frac -= 0.5;
frac *= 2.0; // Show twice as far for each track, because we're doing 2 at once.
}
}
if (TrackProgress(nWhichTrack, frac))
return false;
}
rb.outputLeftTrack->Flush();
if(rightTrack)
rb.outputRightTrack->Flush();
bool bResult =
leftTrack->ClearAndPaste(mCurT0, mCurT1, rb.outputLeftTrack,
true, false, GetTimeWarper());
wxASSERT(bResult); // TO DO: Actually handle this.
if(rightTrack)
{
bResult =
rightTrack->ClearAndPaste(mCurT0, mCurT1, rb.outputRightTrack,
true, false, GetTimeWarper());
wxASSERT(bResult); // TO DO: Actually handle this.
}
}
mCurTrackNum++;
}
else if (mustSync && t->IsSyncLockSelected())
{
t->SyncLockAdjust(mCurT1, mCurT0 + (mCurT1 - mCurT0) * mTotalStretch);
}
//Iterate to the next track
t = iter.Next();
}
if (bGoodResult)
ReplaceProcessedTracks(bGoodResult);
// Update selection
mT0 = mCurT0;
mT1 = mCurT0 + maxDuration;
return bGoodResult;
}
bool EffectNoise::Process()
{
if (noiseDuration <= 0.0)
noiseDuration = sDefaultGenerateLen;
//Iterate over each track
int ntrack = 0;
this->CopyInputWaveTracks(); // Set up mOutputWaveTracks.
bool bGoodResult = true;
#ifdef EXPERIMENTAL_FULL_LINKING
HandleLinkedTracksOnGenerate(noiseDuration, mT0);
#endif
TrackListIterator iter(mOutputWaveTracks);
WaveTrack *track = (WaveTrack *)iter.First();
while (track) {
WaveTrack *tmp = mFactory->NewWaveTrack(track->GetSampleFormat(), track->GetRate());
numSamples = track->TimeToLongSamples(noiseDuration);
sampleCount i = 0;
float *data = new float[tmp->GetMaxBlockSize()];
sampleCount block;
while ((i < numSamples) && bGoodResult) {
block = tmp->GetBestBlockSize(i);
if (block > (numSamples - i))
block = numSamples - i;
MakeNoise(data, block, track->GetRate(), noiseAmplitude);
tmp->Append((samplePtr)data, floatSample, block);
i += block;
//Update the Progress meter
if (TrackProgress(ntrack, (double)i / numSamples))
bGoodResult = false;
}
delete[] data;
tmp->Flush();
track->HandleClear(mT0, mT1, false, false);
track->HandlePaste(mT0, tmp);
delete tmp;
if (!bGoodResult)
break;
//Iterate to the next track
ntrack++;
track = (WaveTrack *)iter.Next();
}
if (bGoodResult)
{
/*
save last used values
save duration unless value was got from selection, so we save only
when user explicitely setup a value
*/
if (mT1 == mT0)
gPrefs->Write(wxT("/CsPresets/NoiseGen_Duration"), noiseDuration);
gPrefs->Write(wxT("/CsPresets/NoiseGen_Type"), noiseType);
gPrefs->Write(wxT("/CsPresets/NoiseGen_Amp"), noiseAmplitude);
mT1 = mT0 + noiseDuration; // Update selection.
}
this->ReplaceProcessedWaveTracks(bGoodResult);
return bGoodResult;
}
int FFmpegImportFileHandle::Import(TrackFactory *trackFactory,
Track ***outTracks,
int *outNumTracks,
Tags *tags)
{
CreateProgress();
// Remove stream contexts which are not marked for importing and adjust mScs and mNumStreams accordingly
for (int i = 0; i < mNumStreams;)
{
if (!mScs[i]->m_use)
{
delete mScs[i];
for (int j = i; j < mNumStreams - 1; j++)
{
mScs[j] = mScs[j+1];
}
mNumStreams--;
}
else i++;
}
mChannels = new WaveTrack **[mNumStreams];
for (int s = 0; s < mNumStreams; s++)
{
switch (mScs[s]->m_stream->codec->sample_fmt)
{
case AV_SAMPLE_FMT_U8:
case AV_SAMPLE_FMT_S16:
case AV_SAMPLE_FMT_U8P:
case AV_SAMPLE_FMT_S16P:
mScs[s]->m_osamplesize = sizeof(int16_t);
mScs[s]->m_osamplefmt = int16Sample;
break;
default:
mScs[s]->m_osamplesize = sizeof(float);
mScs[s]->m_osamplefmt = floatSample;
break;
}
// There is a possibility that number of channels will change over time, but we do not have WaveTracks for new channels. Remember the number of channels and stick to it.
mScs[s]->m_initialchannels = mScs[s]->m_stream->codec->channels;
mChannels[s] = new WaveTrack *[mScs[s]->m_stream->codec->channels];
int c;
for (c = 0; c < mScs[s]->m_stream->codec->channels; c++)
{
mChannels[s][c] = trackFactory->NewWaveTrack(mScs[s]->m_osamplefmt, mScs[s]->m_stream->codec->sample_rate);
if (mScs[s]->m_stream->codec->channels == 2)
{
switch (c)
{
case 0:
mChannels[s][c]->SetChannel(Track::LeftChannel);
mChannels[s][c]->SetLinked(true);
break;
case 1:
mChannels[s][c]->SetChannel(Track::RightChannel);
break;
}
}
else
{
mChannels[s][c]->SetChannel(Track::MonoChannel);
}
}
}
// Handles the start_time by creating silence. This may or may not be correct.
// There is a possibility that we should ignore first N milliseconds of audio instead. I do not know.
/// TODO: Nag FFmpeg devs about start_time until they finally say WHAT is this and HOW to handle it.
for (int s = 0; s < mNumStreams; s++)
{
int64_t stream_delay = 0;
if (mScs[s]->m_stream->start_time != int64_t(AV_NOPTS_VALUE) && mScs[s]->m_stream->start_time > 0)
{
stream_delay = mScs[s]->m_stream->start_time;
wxLogDebug(wxT("Stream %d start_time = %lld, that would be %f milliseconds."), s, (long long) mScs[s]->m_stream->start_time, double(mScs[s]->m_stream->start_time)/AV_TIME_BASE*1000);
}
if (stream_delay != 0)
{
for (int c = 0; c < mScs[s]->m_stream->codec->channels; c++)
{
WaveTrack *t = mChannels[s][c];
t->InsertSilence(0,double(stream_delay)/AV_TIME_BASE);
}
}
}
// This is the heart of the importing process
// The result of Import() to be returend. It will be something other than zero if user canceled or some error appears.
int res = eProgressSuccess;
#ifdef EXPERIMENTAL_OD_FFMPEG
mUsingOD = false;
gPrefs->Read(wxT("/Library/FFmpegOnDemand"), &mUsingOD);
//at this point we know the file is good and that we have to load the number of channels in mScs[s]->m_stream->codec->channels;
//so for OD loading we create the tracks and releasee the modal lock after starting the ODTask.
if (mUsingOD) {
std::vector<ODDecodeFFmpegTask*> tasks;
//append blockfiles to each stream and add an individual ODDecodeTask for each one.
for (int s = 0; s < mNumStreams; s++) {
ODDecodeFFmpegTask* odTask=new ODDecodeFFmpegTask(mScs,mNumStreams,mChannels,mFormatContext, s);
odTask->CreateFileDecoder(mFilename);
//each stream has different duration. We need to know it if seeking is to be allowed.
sampleCount sampleDuration = 0;
if (mScs[s]->m_stream->duration > 0)
sampleDuration = ((sampleCount)mScs[s]->m_stream->duration * mScs[s]->m_stream->time_base.num) *mScs[s]->m_stream->codec->sample_rate / mScs[s]->m_stream->time_base.den;
else
sampleDuration = ((sampleCount)mFormatContext->duration *mScs[s]->m_stream->codec->sample_rate) / AV_TIME_BASE;
// printf(" OD duration samples %qi, sr %d, secs %d\n",sampleDuration, (int)mScs[s]->m_stream->codec->sample_rate,(int)sampleDuration/mScs[s]->m_stream->codec->sample_rate);
//for each wavetrack within the stream add coded blockfiles
for (int c = 0; c < mScs[s]->m_stream->codec->channels; c++) {
WaveTrack *t = mChannels[s][c];
odTask->AddWaveTrack(t);
sampleCount maxBlockSize = t->GetMaxBlockSize();
//use the maximum blockfile size to divide the sections (about 11secs per blockfile at 44.1khz)
for (sampleCount i = 0; i < sampleDuration; i += maxBlockSize) {
sampleCount blockLen = maxBlockSize;
if (i + blockLen > sampleDuration)
blockLen = sampleDuration - i;
t->AppendCoded(mFilename, i, blockLen, c,ODTask::eODFFMPEG);
// This only works well for single streams since we assume
// each stream is of the same duration and channels
res = mProgress->Update(i+sampleDuration*c+ sampleDuration*mScs[s]->m_stream->codec->channels*s,
sampleDuration*mScs[s]->m_stream->codec->channels*mNumStreams);
if (res != eProgressSuccess)
break;
}
}
tasks.push_back(odTask);
}
//Now we add the tasks and let them run, or delete them if the user cancelled
for(int i=0; i < (int)tasks.size(); i++) {
if(res==eProgressSuccess)
ODManager::Instance()->AddNewTask(tasks[i]);
else
{
delete tasks[i];
}
}
} else {
#endif
streamContext *sc = NULL;
// Read next frame.
while ((sc = ReadNextFrame()) != NULL && (res == eProgressSuccess))
{
// ReadNextFrame returns 1 if stream is not to be imported
if (sc != (streamContext*)1)
{
// Decode frame until it is not possible to decode any further
while (sc->m_pktRemainingSiz > 0 && (res == eProgressSuccess || res == eProgressStopped))
{
if (DecodeFrame(sc,false) < 0)
break;
// If something useable was decoded - write it to mChannels
if (sc->m_frameValid)
res = WriteData(sc);
}
// Cleanup after frame decoding
if (sc->m_pktValid)
{
av_free_packet(&sc->m_pkt);
sc->m_pktValid = 0;
}
}
}
// Flush the decoders.
if ((mNumStreams != 0) && (res == eProgressSuccess || res == eProgressStopped))
{
for (int i = 0; i < mNumStreams; i++)
{
if (DecodeFrame(mScs[i], true) == 0)
{
WriteData(mScs[i]);
if (mScs[i]->m_pktValid)
{
av_free_packet(&mScs[i]->m_pkt);
mScs[i]->m_pktValid = 0;
}
}
}
}
#ifdef EXPERIMENTAL_OD_FFMPEG
} // else -- !mUsingOD == true
#endif //EXPERIMENTAL_OD_FFMPEG
// Something bad happened - destroy everything!
if (res == eProgressCancelled || res == eProgressFailed)
{
for (int s = 0; s < mNumStreams; s++)
{
delete[] mChannels[s];
}
delete[] mChannels;
return res;
}
//else if (res == 2), we just stop the decoding as if the file has ended
*outNumTracks = 0;
for (int s = 0; s < mNumStreams; s++)
{
*outNumTracks += mScs[s]->m_initialchannels;
}
// Create new tracks
*outTracks = new Track *[*outNumTracks];
// Copy audio from mChannels to newly created tracks (destroying mChannels elements in process)
int trackindex = 0;
for (int s = 0; s < mNumStreams; s++)
{
for(int c = 0; c < mScs[s]->m_initialchannels; c++)
{
mChannels[s][c]->Flush();
(*outTracks)[trackindex++] = mChannels[s][c];
}
delete[] mChannels[s];
}
delete[] mChannels;
// Save metadata
WriteMetadata(tags);
return res;
}
bool EffectDtmf::Process()
{
if (dtmfDuration <= 0.0)
return false;
//Iterate over each track
this->CopyInputWaveTracks(); // Set up mOutputWaveTracks.
bool bGoodResult = true;
int ntrack = 0;
TrackListIterator iter(mOutputWaveTracks);
WaveTrack *track = (WaveTrack *)iter.First();
while (track) {
// new tmp track, to fill with dtmf sequence
// we will build the track by adding a tone, then a silence, next tone, and so on...
WaveTrack *tmp = mFactory->NewWaveTrack(track->GetSampleFormat(), track->GetRate());
// all dtmf sequence durations in samples from seconds
numSamplesSequence = (longSampleCount)(dtmfDuration * track->GetRate() + 0.5);
numSamplesTone = (longSampleCount)(dtmfTone * track->GetRate() + 0.5);
numSamplesSilence = (longSampleCount)(dtmfSilence * track->GetRate() + 0.5);
// recalculate the sum, and spread the difference - due to approximations.
// Since diff should be in the order of "some" samples, a division (resulting in zero)
// is not sufficient, so we add the additional remaining samples in each tone/silence block,
// at least until available.
int diff = numSamplesSequence - (dtmfNTones*numSamplesTone) - (dtmfNTones-1)*numSamplesSilence;
if (diff>dtmfNTones) {
// in this case, both these values would change, so it makes sense to recalculate diff
// otherwise just keep the value we already have
// should always be the case that dtmfNTones>1, as if 0, we don't even start processing,
// and with 1 there is no difference to spread (no silence slot)...
wxASSERT(dtmfNTones > 1);
numSamplesTone += (diff/(dtmfNTones));
numSamplesSilence += (diff/(dtmfNTones-1));
diff = numSamplesSequence - (dtmfNTones*numSamplesTone) - (dtmfNTones-1)*numSamplesSilence;
}
// this var will be used as extra samples distributor
int extra=0;
longSampleCount i = 0;
longSampleCount j = 0;
int n=0; // pointer to string in dtmfString
sampleCount block;
bool isTone = true;
float *data = new float[tmp->GetMaxBlockSize()];
// for the whole dtmf sequence, we will be generating either tone or silence
// according to a bool value, and this might be done in small chunks of size
// 'block', as a single tone might sometimes be larger than the block
// tone and silence generally have different duration, thus two generation blocks
//
// Note: to overcome a 'clicking' noise introduced by the abrupt transition from/to
// silence, I added a fade in/out of 1/250th of a second (4ms). This can still be
// tweaked but gives excellent results at 44.1kHz: I haven't tried other freqs.
// A problem might be if the tone duration is very short (<10ms)... (?)
//
// One more problem is to deal with the approximations done when calculating the duration
// of both tone and silence: in some cases the final sum might not be same as the initial
// duration. So, to overcome this, we had a redistribution block up, and now we will spread
// the remaining samples in every bin in order to achieve the full duration: test case was
// to generate an 11 tone DTMF sequence, in 4 seconds, and with DutyCycle=75%: after generation
// you ended up with 3.999s or in other units: 3 seconds and 44097 samples.
//
while ((i < numSamplesSequence) && bGoodResult) {
if (isTone)
// generate tone
{
// the statement takes care of extracting one sample from the diff bin and
// adding it into the tone block until depletion
extra=(diff-- > 0?1:0);
for(j=0; j < numSamplesTone+extra; j+=block) {
block = tmp->GetBestBlockSize(j);
if (block > (numSamplesTone+extra - j))
block = numSamplesTone+extra - j;
// generate the tone and append
MakeDtmfTone(data, block, track->GetRate(), dtmfString[n], j, numSamplesTone, dtmfAmplitude);
tmp->Append((samplePtr)data, floatSample, block);
}
i += numSamplesTone;
n++;
if(n>=dtmfNTones)break;
}
else
// generate silence
{
// the statement takes care of extracting one sample from the diff bin and
// adding it into the silence block until depletion
extra=(diff-- > 0?1:0);
for(j=0; j < numSamplesSilence+extra; j+=block) {
block = tmp->GetBestBlockSize(j);
if (block > (numSamplesSilence+extra - j))
block = numSamplesSilence+extra - j;
// generate silence and append
memset(data, 0, sizeof(float)*block);
tmp->Append((samplePtr)data, floatSample, block);
}
i += numSamplesSilence;
}
// flip flag
isTone=!isTone;
//Update the Progress meter
if (TrackProgress(ntrack, (double)i / numSamplesSequence))
bGoodResult = false;
} // finished the whole dtmf sequence
delete[] data;
tmp->Flush();
track->Clear(mT0, mT1);
track->Paste(mT0, tmp);
delete tmp;
if (!bGoodResult)
break;
//Iterate to the next track
ntrack++;
track = (WaveTrack *)iter.Next();
}
if (bGoodResult)
{
/*
save last used values
save duration unless value was got from selection, so we save only
when user explicitely setup a value
*/
if (mT1 == mT0)
gPrefs->Write(wxT("/CsPresets/DtmfGen_SequenceDuration"), dtmfDuration);
gPrefs->Write(wxT("/CsPresets/DtmfGen_String"), dtmfString);
gPrefs->Write(wxT("/CsPresets/DtmfGen_DutyCycle"), dtmfDutyCycle);
gPrefs->Write(wxT("/CsPresets/DtmfGen_Amplitude"), dtmfAmplitude);
// Update selection: this is not accurate if my calculations are wrong.
// To validate, once the effect is done, unselect, and select all, then
// see what the selection length is being reported (in sec,ms,samples)
mT1 = mT0 + dtmfDuration;
}
this->ReplaceProcessedWaveTracks(bGoodResult);
return bGoodResult;
}