1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
|
"""
La Minute Papillon - Mix final (chant + orchestre ambiant moderne)
- Génère un accompagnement orchestral (MIDI) en Dm, 76 bpm, ~4:00
- Rendre le MIDI en WAV via FluidSynth (SoundFont GM)
- Charger la voix (MP3), caler, traitement pro léger, mixer et exporter en WAV
"""
import os
import math
import tempfile
import numpy as np
import soundfile as sf
import librosa
import librosa.effects
from midiutil import MIDIFile
from scipy.signal import butter, lfilter
INPUT_MP3_PATH = "la minute papillon.mp3"
SOUNDFONT_PATH = "/chemin/vers/ton/GM.sf2"
OUTPUT_WAV_PATH = "La_Minute_Papillon_Mix_Final.wav"
TEMPO_BPM = 76
KEY_ROOT_MIDI = 62
TARGET_DURATION = 240.0
INTENSITY_DB = -20.0
ORCH_OFFSET_SEC = 0.0
Dm = [62, 65, 69]; Bb = [58, 62, 65]; F = [53, 57, 60]; C = [55, 59, 62]
Gm = [55, 58, 62]; A7 = [57, 61, 64]
INTRO = [Dm, Bb, F, C]
COUPLET = [Dm, Bb, F, C]
REFRAIN = [Bb, F, Gm, A7]
PONT = [Gm, Dm, Bb, A7]
FINAL = [Dm, Bb, F, Dm]
SECTIONS = [
("Intro", INTRO),
("Couplet", COUPLET),
("Refrain", REFRAIN),
("Couplet", COUPLET),
("Refrain", REFRAIN),
("Pont", PONT),
("Refrain", REFRAIN),
("Final", FINAL),
]
def db_to_lin(db): return 10.0 ** (db / 20.0)
def normalize_peak(y, peak=0.98):
m = np.max(np.abs(y)) + 1e-12
return y * (peak / m)
def butter_highpass(y, sr, cutoff=80.0, order=2):
b, a = butter(order, cutoff / (0.5 * sr), btype='high')
return lfilter(b, a, y)
def simple_comp(y, thresh_db=-14, ratio=2.5, makeup_db=0.0):
rms = np.sqrt(np.mean(y**2) + 1e-12)
lvl_db = 20*np.log10(rms + 1e-12)
if lvl_db > thresh_db:
diff = lvl_db - thresh_db
gain_red_db = diff - diff/ratio
y = y * db_to_lin(-(gain_red_db))
if makeup_db != 0.0:
y *= db_to_lin(makeup_db)
return y
def build_orchestral_midi(midi_path, tempo=TEMPO_BPM):
"""
Pistes:
0 Piano, 1 Flute (lignes discrètes), 2 Strings, 3 Pad, 4 Bass, 5 Percs
Orchestration "ambiant moderne": nappes, arpèges doux, percs vaporeuses.
"""
mf = MIDIFile(6)
for trk in range(6):
mf.addTempo(trk, 0, tempo)
CH_PIANO, CH_FLUTE, CH_STR, CH_PAD, CH_BASS, CH_DR = 0,1,2,3,4,9
PRG_PIANO, PRG_FLUTE, PRG_STR, PRG_PAD, PRG_BASS = 0,73,48,89,32
mf.addProgramChange(0, CH_PIANO, 0, PRG_PIANO)
mf.addProgramChange(1, CH_FLUTE, 0, PRG_FLUTE)
mf.addProgramChange(2, CH_STR, 0, PRG_STR)
mf.addProgramChange(3, CH_PAD, 0, PRG_PAD)
mf.addProgramChange(4, CH_BASS, 0, PRG_BASS)
flute_motif = [62, 64, 65, 67, 69, 70, 72, 74]
t = 0.0
for name, sect in SECTIONS:
for chord in sect:
dur = 4.0
root = min(chord); third = sorted(chord)[1]; fifth = max(chord)
arp = [root, fifth, third, fifth, root, fifth, third, fifth]
for i, n in enumerate(arp):
mf.addNote(0, CH_PIANO, n, t + i*0.5, 0.48, 60)
for n in chord:
mf.addNote(2, CH_STR, n+12, t, dur, 50)
mf.addNote(3, CH_PAD, root-12, t, dur, 38)
mf.addNote(3, CH_PAD, fifth-12, t, dur, 34)
mf.addNote(4, CH_BASS, root-24, t, 2.8, 48)
mf.addNote(4, CH_BASS, root-24, t+2.0, 1.6, 44)
mf.addNote(1, CH_FLUTE, flute_motif[(int(t)) % len(flute_motif)], t+2.5, 1.2, 54)
mf.addNote(5, 9, 42, t+0.0, 0.08, 22)
mf.addNote(5, 9, 37, t+1.5, 0.08, 20)
mf.addNote(5, 9, 42, t+2.0, 0.08, 22)
mf.addNote(5, 9, 44, t+3.0, 0.08, 18)
t += dur
with open(midi_path, "wb") as f:
mf.writeFile(f)
def render_midi_to_wav(midi_path, sf2_path, out_wav_path, sr=48000):
"""Rendu via pyfluidsynth / FluidSynth."""
import fluidsynth
fs = fluidsynth.Synth(samplerate=sr)
sfid = fs.sfload(sf2_path)
fs.program_select(0, sfid, 0, 0)
fs.start(driver="file", filename=out_wav_path)
fs.midi_load(midi_path)
fs.midi_play()
fs.delete()
def main():
assert os.path.exists(INPUT_MP3_PATH), f"Introuvable: {INPUT_MP3_PATH}"
assert os.path.exists(SOUNDFONT_PATH), f"Introuvable: {SOUNDFONT_PATH}"
voice, sr_v = librosa.load(INPUT_MP3_PATH, sr=48000, mono=True)
voice = normalize_peak(voice, 0.98)
with tempfile.TemporaryDirectory() as td:
midi_path = os.path.join(td, "orch.mid")
wav_orch = os.path.join(td, "orch.wav")
build_orchestral_midi(midi_path, tempo=TEMPO_BPM)
render_midi_to_wav(midi_path, SOUNDFONT_PATH, wav_orch, sr=48000)
orch, sr_o = sf.read(wav_orch, dtype="float32")
if orch.ndim == 2:
orch = orch.mean(axis=1)
if sr_o != sr_v:
orch = librosa.resample(orch, orig_sr=sr_o, target_sr=sr_v)
sr_mix = sr_v
else:
sr_mix = sr_o
v = voice.copy()
v = butter_highpass(v, sr_mix, cutoff=80.0, order=2)
v = simple_comp(v, thresh_db=-14, ratio=2.5, makeup_db=0.0)
orch = orch * db_to_lin(INTENSITY_DB)
n = max(len(v), len(orch))
if len(v) < n: v = np.pad(v, (0, n-len(v)))
if len(orch)< n: orch = np.pad(orch, (0, n-len(orch)))
mix = v + orch
mix = normalize_peak(mix, 0.98)
sf.write(OUTPUT_WAV_PATH, mix, sr_mix, subtype="PCM_24")
print(f"✅ Mix final exporté : {OUTPUT_WAV_PATH}")
if __name__ == "__main__":
main()
|
by Randall Munroe, CC BY-NC 2.5
