def fitness(individual, target_features):
"""
Fitness function based on features
Computes the mean squared error between the feature
vectors of two signals
"""
pred_synth = synth.Synth(sr=sr)
pred_params = individual_to_params(individual)
pred_synth.set_parameters(**pred_params)
pred_signal = pred_synth.get_sound_array()
pred_features = extract_features(pred_signal)
return np.mean(np.square(pred_features - target_features)),
def __init__(self, seed=1337):
super().__init__()
self.synth = synth.Synth(sr=44100)
self.r = random.Random(seed)
Licensed under the MIT license.
All text above must be included in any redistribution.
"""
import os
import parser
import sequencer
import synth
import adafruit_trellism4
trellis = adafruit_trellism4.TrellisM4Express(rotation=0)
trellis.pixels.brightness = 0.1
trellis.pixels.fill(0)
syn = synth.Synth()
seq = sequencer.Sequencer(syn)
p = parser.MidiParser()
voices = sorted([
f.split('.')[0] for f in os.listdir('/samples')
if f.endswith('.txt') and not f.startswith('.')
])
print('Voices found: ', voices)
tunes = sorted([
f for f in os.listdir('/midi')
if f.endswith('.mid') and not f.startswith('.')
])
print('Midi files found: ', tunes)
selected_voice = None
import synth as Synth
import numpy as np
import math
import threading
import pyaudio
note_data = json.loads(open('./notes.json').read())
midi_to_note = {}
for note in note_data:
midi_to_note[note_data[note]['midi']] = note
mid = mido.MidiFile('./midi/DancingQueen.mid')
output_io = pyaudio.PyAudio()
S = Synth.Synth(output_io)
for i, msg in enumerate(mid.play()):
try:
midnote = msg.note
_type = msg.type
duration = msg.time
velocity = msg.velocity
if midnote not in midi_to_note:
continue
freq = note_data[midi_to_note[midnote]]['frequency']
if msg.type == 'note_on':
S.play(freq)