def fastAudio(theFile, outFile, chunks: list, speeds: list, log, fps: float,
machineReadable, hideBar):
from wavfile import read, write
import os
import numpy as np
log.checkType(chunks, 'chunks', list)
log.checkType(speeds, 'speeds', list)
def speedsOtherThan1And99999(a: list) -> bool:
return len([x for x in a if x != 1 and x != 99999]) > 0
if (speedsOtherThan1And99999(speeds)):
from audiotsm2 import phasevocoder
from audiotsm2.io.array import ArrReader, ArrWriter
if (len(chunks) == 1 and chunks[0][2] == 0):
log.error('Trying to create an empty file.')
if (not os.path.isfile(theFile)):
log.error('fastAudio.py could not find file: ' + theFile)
samplerate, audioData = read(theFile)
newL = getNewLength(chunks, speeds, fps)
# Get the new length in samples with some extra leeway.
estLeng = int(newL * samplerate * 1.5) + int(samplerate * 2)
# Create an empty array for the new audio.
newAudio = np.zeros((estLeng, 2), dtype=np.int16)
channels = 2
yPointer = 0
audioProgress = ProgressBar(len(chunks), 'Creating new audio',
machineReadable, hideBar)
for chunkNum, chunk in enumerate(chunks):
audioSampleStart = int(chunk[0] / fps * samplerate)
audioSampleEnd = int(audioSampleStart + (samplerate / fps) *
(chunk[1] - chunk[0]))
theSpeed = speeds[chunk[2]]
if (theSpeed != 99999):
spedChunk = audioData[audioSampleStart:audioSampleEnd]
if (theSpeed == 1):
yPointerEnd = yPointer + spedChunk.shape[0]
newAudio[yPointer:yPointerEnd] = spedChunk
else:
spedupAudio = np.zeros((0, 2), dtype=np.int16)
with ArrReader(spedChunk, channels, samplerate, 2) as reader:
with ArrWriter(spedupAudio, channels, samplerate,
2) as writer:
phasevocoder(reader.channels,
speed=theSpeed).run(reader, writer)
spedupAudio = writer.output
yPointerEnd = yPointer + spedupAudio.shape[0]
newAudio[yPointer:yPointerEnd] = spedupAudio
myL = chunk[1] - chunk[0]
mySamples = (myL / fps) * samplerate
newSamples = int(mySamples / theSpeed)
yPointer = yPointer + newSamples
else:
# Speed is too high so skip this section.
yPointerEnd = yPointer
audioProgress.tick(chunkNum)
log.debug('\n - Total Samples: ' + str(yPointer))
log.debug(' - Samples per Frame: ' + str(samplerate / fps))
log.debug(' - Expected video length: ' + str(yPointer /
(samplerate / fps)))
newAudio = newAudio[:yPointer]
write(outFile, samplerate, newAudio)
def fastAudio(ffmpeg, theFile, outFile, silentT, frameMargin, SAMPLE_RATE, audioBit,
verbose, silentSpeed, soundedSpeed, needConvert, chunks=[], fps=30):
if(not os.path.isfile(theFile)):
print('Could not find file:', theFile)
sys.exit(1)
if(outFile == ''):
fileName = theFile[:theFile.rfind('.')]
outFile = f'{fileName}_ALTERED.wav'
if(needConvert):
# Only print this here so other scripts can use this function.
print('Running from fastAudio.py')
import tempfile
from shutil import rmtree
TEMP = tempfile.mkdtemp()
cmd = [ffmpeg, '-i', theFile, '-b:a', audioBit, '-ac', '2', '-ar',
str(SAMPLE_RATE), '-vn', f'{TEMP}/fastAud.wav']
if(not verbose):
cmd.extend(['-nostats', '-loglevel', '0'])
subprocess.call(cmd)
theFile = f'{TEMP}/fastAud.wav'
speeds = [silentSpeed, soundedSpeed]
sampleRate, audioData = read(theFile)
if(chunks == []):
print('Creating chunks')
chunks = getAudioChunks(audioData, sampleRate, fps, silentT, 2, frameMargin)
# Get the estimated length of the new audio in frames.
newL = getNewLength(chunks, speeds, fps)
# Get the new length in samples with some extra leeway.
estLeng = int((newL * sampleRate) * 1.5) + int(sampleRate * 2)
# Create an empty array for the new audio.
newAudio = np.zeros((estLeng, 2), dtype=np.int16)
channels = 2
yPointer = 0
totalChunks = len(chunks)
beginTime = time.time()
for chunkNum, chunk in enumerate(chunks):
audioSampleStart = int(chunk[0] / fps * sampleRate)
audioSampleEnd = int(audioSampleStart + (sampleRate / fps) * (chunk[1] - chunk[0]))
theSpeed = speeds[chunk[2]]
if(theSpeed != 99999):
spedChunk = audioData[audioSampleStart:audioSampleEnd]
spedupAudio = np.zeros((0, 2), dtype=np.int16)
with ArrReader(spedChunk, channels, sampleRate, 2) as reader:
with ArrWriter(spedupAudio, channels, sampleRate, 2) as writer:
phasevocoder(reader.channels, speed=theSpeed).run(
reader, writer
)
spedupAudio = writer.output
yPointerEnd = yPointer + spedupAudio.shape[0]
newAudio[yPointer:yPointerEnd] = spedupAudio
myL = chunk[1] - chunk[0]
mySamples = (myL / fps) * sampleRate
newSamples = int(mySamples / theSpeed)
yPointer = yPointer + newSamples
else:
# Speed is too high so skip this section.
yPointerEnd = yPointer
progressBar(chunkNum, totalChunks, beginTime, title='Creating new audio')
if(verbose):
print('yPointer', yPointer)
print('samples per frame', sampleRate / fps)
print('Expected video length', yPointer / (sampleRate / fps))
newAudio = newAudio[:yPointer]
write(outFile, sampleRate, newAudio)
if('TEMP' in locals()):
rmtree(TEMP)
if(needConvert):
return outFile
def fastAudio(ffmpeg,
theFile,
outFile,
chunks,
speeds,
audioBit,
samplerate,
debug,
needConvert,
log,
fps=30):
if (not os.path.isfile(theFile)):
log.error('Could not find file ' + theFile)
if (needConvert):
# Only print this here so other scripts can use this function.
print('Running from fastAudio.py')
import tempfile
from shutil import rmtree
TEMP = tempfile.mkdtemp()
cmd = [
ffmpeg, '-i', theFile, '-b:a', audioBit, '-ac', '2', '-ar',
str(samplerate), '-vn', f'{TEMP}/fastAud.wav'
]
if (not debug):
cmd.extend(['-nostats', '-loglevel', '0'])
subprocess.call(cmd)
theFile = f'{TEMP}/fastAud.wav'
samplerate, audioData = read(theFile)
newL = getNewLength(chunks, speeds, fps)
# Get the new length in samples with some extra leeway.
estLeng = int(newL * samplerate * 1.5) + int(samplerate * 2)
# Create an empty array for the new audio.
newAudio = np.zeros((estLeng, 2), dtype=np.int16)
channels = 2
yPointer = 0
totalChunks = len(chunks)
beginTime = time.time()
for chunkNum, chunk in enumerate(chunks):
audioSampleStart = int(chunk[0] / fps * samplerate)
audioSampleEnd = int(audioSampleStart + (samplerate / fps) *
(chunk[1] - chunk[0]))
theSpeed = speeds[chunk[2]]
if (theSpeed != 99999):
spedChunk = audioData[audioSampleStart:audioSampleEnd]
if (theSpeed == 1):
yPointerEnd = yPointer + spedChunk.shape[0]
newAudio[yPointer:yPointerEnd] = spedChunk
else:
spedupAudio = np.zeros((0, 2), dtype=np.int16)
with ArrReader(spedChunk, channels, samplerate, 2) as reader:
with ArrWriter(spedupAudio, channels, samplerate,
2) as writer:
phasevocoder(reader.channels,
speed=theSpeed).run(reader, writer)
spedupAudio = writer.output
yPointerEnd = yPointer + spedupAudio.shape[0]
newAudio[yPointer:yPointerEnd] = spedupAudio
myL = chunk[1] - chunk[0]
mySamples = (myL / fps) * samplerate
newSamples = int(mySamples / theSpeed)
yPointer = yPointer + newSamples
else:
# Speed is too high so skip this section.
yPointerEnd = yPointer
progressBar(chunkNum,
totalChunks,
beginTime,
title='Creating new audio')
log.debug('yPointer: ' + str(yPointer))
log.debug('samples per frame: ' + str(samplerate / fps))
log.debug('Expected video length: ' + str(yPointer / (samplerate / fps)))
newAudio = newAudio[:yPointer]
write(outFile, samplerate, newAudio)
if ('TEMP' in locals()):
rmtree(TEMP)
if (needConvert):
conwrite('')
from audiotsm2 import phasevocoder, soundstretch
phasevocoder('sine.wav', 'out.wav', speed=3)
soundstretch('sine.wav', 'out2.wav', speed=3)
def fastAudio(ffmpeg: str, theFile: str, outFile: str, chunks: list,
speeds: list, audioBit, samplerate, needConvert: bool, temp: str,
log, fps: float):
if (len(chunks) == 1 and chunks[0][2] == 0):
log.error('Trying to create empty audio.')
if (not os.path.isfile(theFile)):
log.error('fastAudio.py could not find file: ' + theFile)
if (needConvert):
cmd = [ffmpeg, '-y', '-i', theFile]
if (audioBit is not None):
cmd.extend(['-b:a', str(audioBit)])
cmd.extend(
['-ac', '2', '-ar',
str(samplerate), '-vn', f'{temp}/faAudio.wav'])
if (log.is_ffmpeg):
cmd.extend(['-hide_banner'])
else:
cmd.extend(['-nostats', '-loglevel', '8'])
subprocess.call(cmd)
theFile = f'{temp}/faAudio.wav'
samplerate, audioData = read(theFile)
newL = getNewLength(chunks, speeds, fps)
# Get the new length in samples with some extra leeway.
estLeng = int(newL * samplerate * 1.5) + int(samplerate * 2)
# Create an empty array for the new audio.
newAudio = np.zeros((estLeng, 2), dtype=np.int16)
channels = 2
yPointer = 0
totalChunks = len(chunks)
beginTime = time.time()
for chunkNum, chunk in enumerate(chunks):
audioSampleStart = int(chunk[0] / fps * samplerate)
audioSampleEnd = int(audioSampleStart + (samplerate / fps) *
(chunk[1] - chunk[0]))
theSpeed = speeds[chunk[2]]
if (theSpeed != 99999):
spedChunk = audioData[audioSampleStart:audioSampleEnd]
if (theSpeed == 1):
yPointerEnd = yPointer + spedChunk.shape[0]
newAudio[yPointer:yPointerEnd] = spedChunk
else:
spedupAudio = np.zeros((0, 2), dtype=np.int16)
with ArrReader(spedChunk, channels, samplerate, 2) as reader:
with ArrWriter(spedupAudio, channels, samplerate,
2) as writer:
phasevocoder(reader.channels,
speed=theSpeed).run(reader, writer)
spedupAudio = writer.output
yPointerEnd = yPointer + spedupAudio.shape[0]
newAudio[yPointer:yPointerEnd] = spedupAudio
myL = chunk[1] - chunk[0]
mySamples = (myL / fps) * samplerate
newSamples = int(mySamples / theSpeed)
yPointer = yPointer + newSamples
else:
# Speed is too high so skip this section.
yPointerEnd = yPointer
progressBar(chunkNum,
totalChunks,
beginTime,
title='Creating new audio')
log.debug('\n - Total Samples: ' + str(yPointer))
log.debug(' - Samples per Frame: ' + str(samplerate / fps))
log.debug(' - Expected video length: ' + str(yPointer /
(samplerate / fps)))
newAudio = newAudio[:yPointer]
write(outFile, samplerate, newAudio)
if (needConvert):
conwrite('')