def chord_translate(chord_str: str,
scale: int,
base_oct: int = 3) -> List[str]:
chord_str = chord_str.replace("(", "").replace(")", "")
try:
chord = pychord.Chord(chord_str)
addomit = dict()
except ValueError as e:
# add omit だけ抜き取りたい
chord_without_quality = re.sub(f"^{note}", "", chord_str)
valid_qualities = list(pychord.quality.QUALITY_DICT.keys())
valid_qualities.remove("")
valid_qualities.sort(key=len, reverse=True)
found = False
addomit_str = ""
for quality in valid_qualities:
if chord_without_quality.find(quality) == 0: # 先頭のコードqualityを削除する
found = True
addomit_str = re.sub(f"^{re.escape(quality)}", "",
chord_without_quality)
chord = pychord.Chord(
re.sub(f"{re.escape(addomit_str)}$", "", chord_str))
break
if not found:
raise e
# TODO: Fix this
# Should raise error if removed addomit_str is not empty
addomit = {
"adds":
[r.groups()[-2:] for r in re.finditer(add_regex, addomit_str)],
"omits":
[r.groups()[-1] for r in re.finditer(omit_regex, addomit_str)],
}
if len(addomit) > 0:
s = musthe.Scale(f"{chord.root}3", "major")
adds = addomit.get("adds")
omits = addomit.get("omits")
for add in adds:
align = (add[0].count("#") + add[0].count("+") -
add[0].count("b") - add[0].count("-"))
adder_note_dict = parse_note_str(
transpose(s[int(add[1]) - 1].scientific_notation(), align))
chord.quality.append_note(adder_note_dict["note"], chord._root,
int(adder_note_dict["oct"]) - 3)
for omit in omits:
for o in range(int(omit) - 2, int(omit) + 1):
if o in chord.quality.components:
chord.quality.components.remove(o)
chord.transpose(scale)
notes = chord.components_with_pitch(base_oct)
return notes
def generate_harmony(matches, next_matches, initial):
global track_hypothesis, pos_hypothesis
# print(f"Current hypothesis: {track_hypothesis}:{pos_hypothesis}")
track_matches = matches[matches.track == track_hypothesis]
if not track_matches.empty:
pos_matches = track_matches[track_matches.position == pos_hypothesis]
if pos_matches.empty:
prev_match = track_matches.iloc[0]
else:
prev_match = pos_matches.iloc[0]
result = next_matches[next_matches.track.isin([prev_match.track]) &
next_matches.position.isin([prev_match.position +
c.N_NGRAM])].iloc[0]
pos_hypothesis = result.position
else:
prev_match = matches.iloc[0]
result = next_matches.iloc[0]
track_hypothesis = result.track
pos_hypothesis = result.position
duration = result.duration.split()
harmony = [pychord.Chord(h) for h in result.harmony.split() if h != 'None']
delta = (initial - prev_match.initial) % 12
for i in range(len(harmony)):
harmony[i].transpose(delta)
return harmony, duration
def parse_song(text):
chords = []
for token in text.split():
try:
chords.append(pychord.Chord(token))
except ValueError:
pass
return chords
def get_user_chord_progression(self) -> List:
"""Calls for user input and returns a list of chord objects.
"""
userInput = input()
while userInput[-1] != '.':
if userInput == 'q':
print('~ Goodbye! ~')
sys.exit()
print('Oops! Make sure to include a period (.) at the end.')
sys.stdout.flush()
userInput = input()
# remove the trailing period
userInput = userInput[:-1]
# remove all whitespace for each entry
chordProgression = [
pc.Chord(''.join(s.strip())) for s in userInput.split(',')
]
return chordProgression
txt = open(args.txt)
time.sleep(1)
current_chord = None
for line in txt:
note, duration, chord = line.split()
duration = float(fractions.Fraction(duration))
if args.c and chord != 'None' and chord != current_chord:
if args.x:
print(f'CHORD {chord} {datetime.now().time()}')
current_chord = chord
if args.s:
threading.Thread(target=play.play_chord_async,
args=(pychord.Chord(current_chord), 3)).start()
if note[0] == 'R':
time.sleep(util.duration_to_sec(duration, bpm))
else:
if args.x:
print(f'NOTE {util.canonical_note(note)} {datetime.now().time()}')
if args.n:
keyboard.press(c.note_to_key[note])
time.sleep(util.duration_to_sec(duration, bpm))
if args.n:
keyboard.release(c.note_to_key[note])
with mido.open_output() as outport:
outport.send(mido.Message('stop'))
def toPychord(self):
"""
Convert this Chord object to a pychord.Chord.
If this Chord is key irrespective, it is put into the key of C
"""
chord_root = self.chord[:self.chord.find(":")]
find_bass = self.chord.find("/")
if find_bass == -1:
chord_quality = self.chord[self.chord.find(":") + 1:]
chord_bass = ''
else:
chord_quality = self.chord[self.chord.find(":") +
1:self.chord.find("/")]
chord_bass = self.chord[self.chord.find("/") + 1:]
# Create notes list to reference later. Only sharps are used. Any flats can look to one index before their base.
notes = [
"A", "A#", "B", "C", "C#", "D", "D#", "E", "F", "F#", "G", "G#"
]
# Record intervals for a generic scale (2 = whole step, 1 = half step)
intervals = [2, 2, 1, 2, 2, 2, 1]
# If key_irrespective, convert to key of C
if self.key_irrespective:
# Check for 'x:__' chords and return None
if chord_root == 'x':
return None
# Check for 'b' in chord_root
offset = 0
if chord_root.find('b') != -1:
offset = -1
chord_root = chord_root[0]
chord_root = int(chord_root) - 1 # index from 0 instead of 1
chord_index = 3 # Begin at C
for i in range(chord_root):
chord_index += intervals[i]
chord_index += offset
chord_index %= len(notes)
chord_root = notes[chord_index]
# Definining quality conversions based on http://ismir2005.ismir.net/proceedings/1080.pdf
# and https://github.com/yuma-m/pychord/blob/master/pychord/constants/qualities.py
# reveals that 'maj' => 'M' if not alone and 'min' => 'm' if not alone are the only
# required conversions
new_quality = chord_quality
"""if new_quality.find('maj') != -1 and new_quality != 'maj':
new_quality = new_quality.replace('maj', 'M')
if new_quality.find('min') != -1 and new_quality != 'min':
new_quality = new_quality.replace('min', 'm')
if new_quality.find('hdim7') != -1:
new_quality = new_quality.replace('hdim7', 'm7-5')
if new_quality.find('(9)') != -1:
new_quality = new_quality.replace('(9)', '9')
if new_quality.find('(11)') != -1:
new_quality = new_quality.replace('(11)', '')
if new_quality == 'M':
new_quality = new_quality.replace('M', 'maj')
if new_quality == 'M6':
new_quality = new_quality.replace('M6', '6')
if new_quality.find('6'):
new_quality = new_quality.replace('6', '')
if new_quality == 'M(b7)':
new_quality = new_quality.replace('M(b7)', 'm7')"""
if 'maj' in new_quality and new_quality != 'maj':
# Replace 'maj' with 'M' for 7th chords
new_quality = new_quality.replace('maj', 'M')
if 'min' in new_quality and new_quality != 'min':
# Replace 'min' with 'm' for 7th chords
new_quality = new_quality.replace('min', 'm')
if 'hdim7' in new_quality:
new_quality = new_quality.replace('hdim7', 'm7-5')
# Remove parenthesis from qualities
if '(' in new_quality:
new_quality = new_quality.replace('(', '')
if ')' in new_quality:
new_quality = new_quality.replace(')', '')
# Create pychord.Chord with converted values (excluding bass)
try:
result_chord = pychord.Chord(chord_root + new_quality)
except:
if 'maj' in new_quality or 'M' in new_quality:
new_quality = 'maj'
elif 'min' in new_quality or 'm' in new_quality:
new_quality = 'min'
try:
result_chord = pychord.Chord(chord_root + new_quality)
except:
# Something is funky with the formatting
return None
# If no bass or unreadable bass, return un-inverted chord
try:
int(chord_bass)
except:
return result_chord
# Configure bass
chord_notes = list(result_chord.components())
try:
result_chord = pychord.Chord(chord_root + new_quality + "/" +
chord_notes[int(chord_bass) - 3])
except:
pass
return result_chord
def handle(self, argv=None):
"""
Main function.
Parses command, load settings and dispatches accordingly.
"""
help_message = "Please supply chord progression!. See --help for more options."
parser = argparse.ArgumentParser(
description=
'chords2midi - Create MIDI files from written chord progressions.\n'
)
parser.add_argument('progression',
metavar='U',
type=str,
nargs='*',
help=help_message)
parser.add_argument('-B',
'--bassline',
action='store_true',
default=False,
help='Throw an extra bassline on the pattern')
parser.add_argument('-b',
'--bpm',
type=int,
default=80,
help='Set the BPM (default 80)')
parser.add_argument('-t',
'--octave',
type=str,
default='4',
help='Set the octave(s) (ex: 3,4) (default 4)')
parser.add_argument('-i',
'--input',
type=str,
default=None,
help='Read from an input file.')
parser.add_argument('-k',
'--key',
type=str,
default='C',
help='Set the key (default C)')
parser.add_argument('-n',
'--notes',
type=int,
default=99,
help='Notes in each chord (default all)')
parser.add_argument('-d',
'--duration',
type=float,
default=1.0,
help='Set the chord duraction (default 1)')
parser.add_argument(
'-D',
'--directory',
action='store_true',
default=False,
help=
'Output the contents to the directory of the input progression.')
parser.add_argument(
'-H',
'--humanize',
type=float,
default=0.0,
help=
'Set the amount to "humanize" (strum) a chord, in ticks - try .11 (default 0.0)'
)
parser.add_argument(
'-o',
'--output',
type=str,
help=
'Set the output file path. Default is the current key and progression in the current location.'
)
parser.add_argument(
'-O',
'--offset',
type=float,
default=0.0,
help='Set the amount to offset each chord, in ticks. (default 0.0)'
)
parser.add_argument('-p',
'--pattern',
type=str,
default=None,
help='Set the pattern. Available patterns: ' +
(', '.join(patterns.keys())))
parser.add_argument(
'-r',
'--reverse',
action='store_true',
default=False,
help='Reverse a progression from C-D-E format into I-II-III format'
)
parser.add_argument('-v',
'--version',
action='store_true',
default=False,
help='Display the current version of chords2midi')
args = parser.parse_args(argv)
self.vargs = vars(args)
if self.vargs['version']:
version = pkg_resources.require("chords2midi")[0].version
print(version)
return
# Support `c2m I III V and `c2m I,III,V` formats.
if not self.vargs['input']:
if len(self.vargs['progression']) < 1:
print("You need to supply a progression! (ex I V vi IV)")
return
if len(self.vargs['progression']) < 2:
progression = self.vargs['progression'][0].split(',')
else:
progression = self.vargs['progression']
else:
with open(self.vargs['input']) as fn:
content = ''.join(fn.readlines()).strip()
content = content.replace('\n', ' ').replace(',', ' ')
progression = content.split(' ')
og_progression = progression
# If we're reversing, we don't need any of the MIDI stuff.
if self.vargs['reverse']:
result = ""
key = self.vargs['key']
for item in progression:
comps = pychord.Chord(item).components()
position = determine(comps, key, True)[0]
if 'M' in position:
position = position.upper()
position = position.replace('M', '')
if 'm' in position:
position = position.lower()
position = position.replace('m', '')
if 'B' in position:
position = position + "b"
position = position.replace('B', '')
result = result + position + " "
print result
return
track = 0
channel = 0
ttime = 0
duration = self.vargs['duration'] # In beats
tempo = self.vargs['bpm'] # In BPM
volume = 100 # 0-127, as per the MIDI standard
bar = 0
humanize_interval = self.vargs['humanize']
directory = self.vargs['directory']
num_notes = self.vargs['notes']
offset = self.vargs['offset']
key = self.vargs['key']
octaves = self.vargs['octave'].split(',')
root_lowest = self.vargs.get('root_lowest', False)
bassline = self.vargs['bassline']
pattern = self.vargs['pattern']
# Could be interesting to do multiple parts at once.
midi = MIDIFile(1)
midi.addTempo(track, ttime, tempo)
##
# Main generator
##
has_number = False
progression_chords = []
# Apply patterns
if pattern:
if pattern not in patterns.keys():
print("Invalid pattern! Must be one of: " +
(', '.join(patterns.keys())))
return
new_progression = []
input_progression = progression[:] # 2.7 copy
pattern_mask = patterns[pattern]
pattern_mask_index = 0
current_chord = None
while True:
pattern_instruction = pattern_mask[pattern_mask_index]
if pattern_instruction == "N":
if len(input_progression) == 0:
break
current_chord = input_progression.pop(0)
new_progression.append(current_chord)
elif pattern_instruction == "S":
new_progression.append(current_chord)
elif pattern_instruction == "X":
new_progression.append("X")
if pattern_mask_index == len(pattern_mask) - 1:
pattern_mask_index = 0
else:
pattern_mask_index = pattern_mask_index + 1
progression = new_progression
# We do this to allow blank spaces
for chord in progression:
# This is for # 'I', 'VI', etc
progression_chord = to_chords(chord, key)
if progression_chord != []:
has_number = True
# This is for 'C', 'Am', etc.
if progression_chord == []:
try:
progression_chord = [pychord.Chord(chord).components()]
except Exception:
# This is an 'X' input
progression_chord = [None]
chord_info = {}
chord_info['notes'] = progression_chord[0]
if has_number:
chord_info['number'] = chord
else:
chord_info['name'] = chord
if progression_chord[0]:
chord_info['root'] = progression_chord[0][0]
else:
chord_info['root'] = None
progression_chords.append(chord_info)
# For each input..
previous_pitches = []
for chord_index, chord_info in enumerate(progression_chords):
# Unpack object
chord = chord_info['notes']
# NO_OP
if chord == None:
bar = bar + 1
continue
root = chord_info['root']
root_pitch = pychord.utils.note_to_val(
notes.int_to_note(notes.note_to_int(root)))
# Reset internals
humanize_amount = humanize_interval
pitches = []
all_new_pitches = []
# Turns out this algorithm was already written in the 1800s!
# https://en.wikipedia.org/wiki/Voice_leading#Common-practice_conventions_and_pedagogy
# a) When a chord contains one or more notes that will be reused in the chords immediately following, then these notes should remain, that is retained in the respective parts.
# b) The parts which do not remain, follow the law of the shortest way (Gesetze des nachsten Weges), that is that each such part names the note of the following chord closest to itself if no forbidden succession XXX GOOD NAME FOR A BAND XXX arises from this.
# c) If no note at all is present in a chord which can be reused in the chord immediately following, one must apply contrary motion according to the law of the shortest way, that is, if the root progresses upwards, the accompanying parts must move downwards, or inversely, if the root progresses downwards, the other parts move upwards and, in both cases, to the note of the following chord closest to them.
root = None
for i, note in enumerate(chord):
# Sanitize notes
sanitized_notes = notes.int_to_note(notes.note_to_int(note))
pitch = pychord.utils.note_to_val(sanitized_notes)
if i == 0:
root = pitch
if root:
if root_lowest and pitch < root: # or chord_index is 0:
pitch = pitch + 12 # Start with the root lowest
all_new_pitches.append(pitch)
# Reuse notes
if pitch in previous_pitches:
pitches.append(pitch)
no_melodic_fluency = False # XXX: vargify
if previous_pitches == [] or all_new_pitches == [] or pitches == [] or no_melodic_fluency:
pitches = all_new_pitches
else:
# Detect the root direction
root_upwards = None
if pitches[0] >= all_new_pitches[0]:
root_upwards = True
else:
root_upwards = False
# Move the shortest distance
if pitches != []:
new_remaining_pitches = list(all_new_pitches)
old_remaining_pitches = list(previous_pitches)
for i, new_pitch in enumerate(all_new_pitches):
# We're already there
if new_pitch in pitches:
new_remaining_pitches.remove(new_pitch)
old_remaining_pitches.remove(new_pitch)
continue
# Okay, so need to find the overall shortest distance from the remaining pitches - including their permutations!
while len(new_remaining_pitches) > 0:
nearest_distance = 9999
previous_index = None
new_index = None
pitch_to_add = None
for i, pitch in enumerate(new_remaining_pitches):
# XXX: DRY
# The Pitch
pitch_to_test = pitch
nearest = min(old_remaining_pitches,
key=lambda x: abs(x - pitch_to_test))
old_nearest_index = old_remaining_pitches.index(
nearest)
if nearest < nearest_distance:
nearest_distance = nearest
previous_index = old_nearest_index
new_index = i
pitch_to_add = pitch_to_test
# +12
pitch_to_test = pitch + 12
nearest = min(old_remaining_pitches,
key=lambda x: abs(x - pitch_to_test))
old_nearest_index = old_remaining_pitches.index(
nearest)
if nearest < nearest_distance:
nearest_distance = nearest
previous_index = old_nearest_index
new_index = i
pitch_to_add = pitch_to_test
# -12
pitch_to_test = pitch - 12
nearest = min(old_remaining_pitches,
key=lambda x: abs(x - pitch_to_test))
old_nearest_index = old_remaining_pitches.index(
nearest)
if nearest < nearest_distance:
nearest_distance = nearest
previous_index = old_nearest_index
new_index = i
pitch_to_add = pitch_to_test
# Before we add it - just make sure that there isn't a better place for it.
pitches.append(pitch_to_add)
del old_remaining_pitches[previous_index]
del new_remaining_pitches[new_index]
# This is for the C E7 type scenario
if len(old_remaining_pitches) == 0:
for x, extra_pitch in enumerate(
new_remaining_pitches):
pitches.append(extra_pitch)
del new_remaining_pitches[x]
# Final check - can the highest and lowest be safely folded inside?
max_pitch = max(pitches)
min_pitch = min(pitches)
index_max = pitches.index(max_pitch)
folded_max = max_pitch - 12
if (folded_max > min_pitch) and (folded_max
not in pitches):
pitches[index_max] = folded_max
max_pitch = max(pitches)
min_pitch = min(pitches)
index_min = pitches.index(min_pitch)
folded_min = min_pitch + 12
if (folded_min < max_pitch) and (folded_min
not in pitches):
pitches[index_min] = folded_min
# Make sure the average can't be improved
# XXX: DRY
if len(previous_pitches) != 0:
previous_average = sum(previous_pitches) / len(
previous_pitches)
# Max
max_pitch = max(pitches)
min_pitch = min(pitches)
index_max = pitches.index(max_pitch)
folded_max = max_pitch - 12
current_average = sum(pitches) / len(pitches)
hypothetical_pitches = list(pitches)
hypothetical_pitches[index_max] = folded_max
hypothetical_average = sum(hypothetical_pitches) / len(
hypothetical_pitches)
if abs(previous_average -
hypothetical_average) <= abs(previous_average -
current_average):
pitches[index_max] = folded_max
# Min
max_pitch = max(pitches)
min_pitch = min(pitches)
index_min = pitches.index(min_pitch)
folded_min = min_pitch + 12
current_average = sum(pitches) / len(pitches)
hypothetical_pitches = list(pitches)
hypothetical_pitches[index_min] = folded_min
hypothetical_average = sum(hypothetical_pitches) / len(
hypothetical_pitches)
if abs(previous_average -
hypothetical_average) <= abs(previous_average -
current_average):
pitches[index_min] = folded_min
# Apply contrary motion
else:
print("Applying contrary motion!")
for i, new_pitch in enumerate(all_new_pitches):
if i == 0:
pitches.append(new_pitch)
continue
# Root upwards, the rest move down.
if root_upwards:
if new_pitch < previous_pitches[i]:
pitches.append(new_pitch)
else:
pitches.append(new_pitch - 12)
else:
if new_pitch > previous_pitches[i]:
pitches.append(new_pitch)
else:
pitches.append(new_pitch + 12)
# Bassline
if bassline:
pitches.append(root_pitch - 24)
# Melody
# Octave is a simple MIDI offset counter
for octave in octaves:
for note in pitches:
pitch = int(note) + (int(octave.strip()) * 12)
# Don't humanize bassline note
if bassline and (pitches.index(note) == len(pitches) - 1):
midi_time = offset + bar
else:
midi_time = offset + bar + humanize_amount
# Write the note
midi.addNote(track=track,
channel=channel,
pitch=pitch,
time=midi_time,
duration=duration,
volume=volume)
humanize_amount = humanize_amount + humanize_interval
if i + 1 >= num_notes:
break
bar = bar + 1
previous_pitches = pitches
##
# Output
##
if self.vargs['output']:
filename = self.vargs['output']
elif self.vargs['input']:
filename = self.vargs['input'].replace('.txt', '.mid')
else:
if has_number:
key_prefix = key + '-'
else:
key_prefix = ''
filename = key_prefix + '-'.join(og_progression) + '-' + str(tempo)
if bassline:
filename = filename + "-bassline"
if pattern:
filename = filename + "-" + pattern
if os.path.exists(filename):
filename = key_prefix + '-'.join(og_progression) + '-' + str(
tempo) + '-' + str(int(time.time()))
filename = filename + '.mid'
if directory:
directory_to_create = '-'.join(og_progression)
try:
os.makedirs(directory_to_create)
except OSError as exc: # Python >2.5
if exc.errno == errno.EEXIST and os.path.isdir(
directory_to_create):
pass
else:
raise
filename = directory_to_create + '/' + filename
with open(filename, "wb") as output_file:
midi.writeFile(output_file)
def harmonic_random(reffile):
chords = []
lf = open(reffile)
nf = open(reffile)
for line in lf:
label, value, time = line.split()
if random.randint(1,10) <= 3:
continue
time = datetime.strptime(time, '%H:%M:%S.%f')
chord = gen_rand_chord()
chords.append((pychord.Chord(chord), time))
chord_groups = [[] for c in chords]
for line in nf:
label, value, time = line.split()
if label != 'NOTE':
continue
time = datetime.strptime(time, '%H:%M:%S.%f')
if time < chords[0][1]:
continue
for i in range(len(chords)-1):
if chords[i][1] <= time < chords[i+1][1]:
chord_groups[i].append(value[:-1])
break
else:
assert chords[-1][1] <= time
chord_groups[-1].append(value[:-1])
bad_notes = 0
total_notes = 0
for i in range(len(chords)):
chord = chords[i][0]
notes = chord_groups[i]
quality = chord.quality.quality
if quality in major_quals:
quality = 'major'
elif quality in minor_quals:
quality = 'natural_minor'
else:
quality = None
if quality:
scale = musthe.Scale(musthe.Note(chord.root), quality)
scale_notes = [str(scale[i]) for i in range(len(scale))]
else:
scale = []
for i in range(len(notes)):
lower = max(0, i-HARMONIC_WINDOW)
upper = min(len(notes)-1, i+HARMONIC_WINDOW)
relevant_notes = [notes[j] for j in range(lower, upper+1)]
for note in relevant_notes:
if note in chord.components():# or musthe.Note(note) in scale:
break
else:
bad_notes += 1
total_notes += len(notes)
# print('{:>8}'.format(f'{bad_notes}/{total_notes}\t'), end='')
# print(f'{bad_notes/total_notes*100:.0f}% error')
return bad_notes / total_notes