Using AKSampleDescriptor - ios

Question

Using AKSamplerDescriptor
I am using an adapted AKSampler example, in which I try to use the sforzando output of Fluid.sf3 melodicSounds. Sforzando creates .sfz files for each instrument, but all pointing for the global sample to a huge .wav file.
In all the instrument.sfz files there is an offset and endpoint description for the part of the wave file to be used.
When I load the .sfz file I get a crash due to memory problems. It seems that for every defined region in the .sfz file the complete .wav file (140 mB) is loaded again.
The most likely is that loading the sample file with the AKSampleDescriptor as done in the AKSampler example will ignore offset and endpoint (AKSampleDescriptor.startPoint and AKSampleDescriptor.endPoint) while reloading the complete .wav file.
Is there a way to load just the part start-to-end wanted from the sample file, because the complete file has al the sample data for all the instruments (I know and use polyphony that extracts only one instrument at the time and works fine, but this is for other use)
Or, and that seems the best to me, just load the file once and than have the sampledescriptors point to the data in memory

Answer 1

Good suggestions, Rob. I just ran into this one-giant-WAV issue myself, having never seen it before. I was also using Sforzando for conversion. I'll look into adding the necessary capabilities to AKSampler. In the meantime, it might be easier to write a program to cut up the one WAV file into smaller pieces and adjust the SFZ accordingly.
Here is some Python 2.7 code to do this, which I have used successfully with a Sforzando-converted sf2 soundfont. It might need changes to work for you--there is huge variability among sfz files--but at least it might help you get started. This code requires the PyDub library for manipulating WAV audio.
import os
import re
from pydub import AudioSegment
def stripComments(text):
def replacer(match):
s = match.group(0)
if s.startswith('/'):
return " " # note: a space and not an empty string
else:
return s
pattern = re.compile(
r'//.*?$|/\*.*?\*/|\'(?:\\.|[^\\\'])*\'|"(?:\\.|[^\\"])*"',
re.DOTALL | re.MULTILINE
)
return re.sub(pattern, replacer, text)
def updateSplitList(splitList, regionLabels, values):
if len(values) > 3:
start = int(values['offset'])
length = int(values['end']) - start
name = regionLabels.pop(0)
splitList.add((name, start, length))
def lookupSplitName(splitList, offset, end):
for (name, start, end) in splitList:
if offset == start and end == end:
return name
return None
def outputGroupAndRegion(outputFile, splitList, values):
if values.has_key('lokey') and values.has_key('hikey') and values.has_key('pitch_keycenter'):
outputFile.write('<group> lokey=%s hikey=%s pitch_keycenter=%s\n' % (values['lokey'], values['hikey'], values['pitch_keycenter']))
elif values.has_key('key') and values.has_key('pitch_keycenter'):
outputFile.write('<group> key=%s pitch_keycenter=%s\n' % (values['key'], values['pitch_keycenter']))
if len(values) > 3:
outputFile.write(' <region> ')
if values.has_key('lovel') and values.has_key('hivel'):
outputFile.write('lovel=%s hivel=%s ' % (values['lovel'], values['hivel']))
if values.has_key('tune'):
outputFile.write('tune=%s ' % values['tune'])
if values.has_key('volume'):
outputFile.write('volume=%s ' % values['volume'])
if values.has_key('offset'):
outputFile.write('offset=0 ')
if values.has_key('end'):
outputFile.write('end=%d ' % (int(values['end']) - int(values['offset'])))
if values.has_key('loop_mode'):
outputFile.write('loop_mode=%s ' % values['loop_mode'])
if values.has_key('loop_start'):
outputFile.write('loop_start=%d ' % (int(values['loop_start']) - int(values['offset'])))
if values.has_key('loop_end'):
outputFile.write('loop_end=%d ' % (int(values['loop_end']) - int(values['offset'])))
outputFile.write('sample=samples/%s' % lookupSplitName(splitList, int(values['offset']), int(values['end'])) + '.wav\n')
def process(inputFile, outputFile):
# create a list of region labels
regionLabels = list()
for line in open(inputFile):
if line.strip().startswith('region_label'):
regionLabels.append(line.strip().split('=')[1])
# read entire input SFZ file
sfz = open(inputFile).read()
# strip comments and create a mixed list of <header> tags and key=value pairs
sfz_list = stripComments(sfz).split()
inSection = "none"
default_path = ""
global_sample = None
values = dict()
splitList = set()
# parse the input SFZ data and build up splitList
for item in sfz_list:
if item.startswith('<'):
inSection = item
updateSplitList(splitList, regionLabels, values)
values.clear()
continue
elif item.find('=') < 0:
#print 'unknown:', item
continue
key, value = item.split('=')
if inSection == '<control>' and key == 'default_path':
default_path = value.replace('\\', '/')
elif inSection == '<global>' and key == 'sample':
global_sample = value.replace('\\', '/')
elif inSection == '<region>':
values[key] = value
# split the wav file
bigWav = AudioSegment.from_wav(global_sample)
#print "%d channels, %d bytes/sample, %d frames/sec" % (bigWav.channels, bigWav.sample_width, bigWav.frame_rate)
frate = float(bigWav.frame_rate)
for (name, start, length) in splitList:
startMs = 1000 * start / frate
endMs = 1000 * (start + length) / frate
wav = bigWav[startMs : endMs]
wavName = 'samples/' + name + '.wav'
wav.export(wavName, format='wav')
# parse the input SFZ data again and generate the output SFZ
for item in sfz_list:
if item.startswith('<'):
inSection = item
outputGroupAndRegion(outputFile, splitList, values)
values.clear()
continue
elif item.find('=') < 0:
#print 'unknown:', item
continue
key, value = item.split('=')
if inSection == '<control>' and key == 'default_path':
default_path = value.replace('\\', '/')
elif inSection == '<global>' and key == 'sample':
global_sample = value.replace('\\', '/')
elif inSection == '<region>':
values[key] = value
dirPath = '000'
fileNameList = os.listdir(dirPath)
for fileName in fileNameList:
if fileName.endswith('.sfz'):
inputFile = os.path.join(dirPath, fileName)
outputFile = open(fileName, 'w')
print fileName
process(inputFile, outputFile)