-
Notifications
You must be signed in to change notification settings - Fork 61
/
Copy pathJoin_the_dots.py
178 lines (170 loc) · 7.57 KB
/
Join_the_dots.py
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
import pickle
import numpy as np
from mutagen.mp3 import MP3
from mutagen.mp4 import MP4
from mutagen.id3 import ID3, ID3NoHeaderError
import subprocess as sp
import random
import argparse
import re
max_duration = 10 * 60 # avoid adding mixes to mix
def get_track_duration(filename):
duration = 0
if filename[-3:].lower() == 'mp3':
duration = MP3(filename).info.length
elif filename[-3:].lower() == 'm4a':
duration = MP4(filename).info.length
return duration
def most_similar(positive=[], negative=[], topn=5, noise=0):
if isinstance(positive, str):
positive = [positive] # broadcast to list
if isinstance(negative, str):
negative = [negative] # broadcast to list
mp3_vec_i = np.sum([mp3tovec[i] for i in positive] + [-mp3tovec[i] for i in negative], axis=0)
mp3_vec_i += np.random.normal(0, noise * np.linalg.norm(mp3_vec_i), len(mp3_vec_i))
mp3_vec_i_norm = np.linalg.norm(mp3_vec_i) # precalculate norms for mp3_vec_i
similar = []
for track_j in mp3tovec:
if track_j in positive or track_j in negative:
continue
mp3_vec_j = mp3tovec[track_j]
mp3_vec_j_norm = mp3_vec_j_norms[track_j]
cos_proximity = np.dot(mp3_vec_i, mp3_vec_j) / (mp3_vec_i_norm * mp3_vec_j_norm)
similar.append((track_j, cos_proximity))
return sorted(similar, key=lambda x:-x[1])[:topn]
def most_similar_by_vec(positive=[], negative=[], topn=5, noise=0):
if isinstance(positive, str):
positive = [positive] # broadcast to list
if isinstance(negative, str):
negative = [negative] # broadcast to list
mp3_vec_i = np.sum([i for i in positive] + [-i for i in negative], axis=0)
mp3_vec_i += np.random.normal(0, noise * np.linalg.norm(mp3_vec_i), len(mp3_vec_i))
mp3_vec_i_norm = np.linalg.norm(mp3_vec_i) # precalculate norms for mp3_vec_i
similar = []
for track_j in mp3tovec:
mp3_vec_j = mp3tovec[track_j]
mp3_vec_j_norm = mp3_vec_j_norms[track_j]
cos_proximity = np.dot(mp3_vec_i, mp3_vec_j) / (mp3_vec_i_norm * mp3_vec_j_norm)
similar.append((track_j, cos_proximity))
return sorted(similar, key=lambda x:-x[1])[:topn]
def in_playlist(candidate, playlist):
if candidate in playlist:
return True
for track in playlist:
mp3_vec_i = mp3tovec[candidate]
mp3_vec_i_norm = np.linalg.norm(mp3_vec_i)
mp3_vec_j = mp3tovec[track]
mp3_vec_j_norm = np.linalg.norm(mp3_vec_j)
cos_proximity = np.dot(mp3_vec_i, mp3_vec_j) / (mp3_vec_i_norm * mp3_vec_j_norm)
if cos_proximity > 1 - epsilon_distance:
return True
return False
def make_playlist(seed_tracks, size=10, lookback=3, noise=0):
max_tries = 100
playlist = seed_tracks
while len(playlist) < size:
similar = most_similar(positive=playlist[-lookback:], topn=max_tries, noise=noise)
candidates = [candidate[0] for candidate in similar if candidate[0] != playlist[-1]]
for candidate in candidates:
if not in_playlist(candidate, playlist) and get_track_duration(candidate) < max_duration:
break
playlist.append(candidate)
return playlist
def join_the_dots(tracks, n=5, noise=0): # create a musical journey between given track "waypoints"
max_tries = 100
playlist = []
end = start = tracks[0]
start_vec = mp3tovec[start]
for end in tracks[1:]:
end_vec = mp3tovec[end]
playlist.append(start)
for i in range(n):
similar = most_similar_by_vec(positive=[(n-i+1)/n * start_vec + (i+1)/n * end_vec], topn=max_tries, noise=noise)
candidates = [candidate[0] for candidate in similar if candidate[0] != playlist[-1]]
for candidate in candidates:
if not in_playlist(candidate, playlist) and candidate != end and get_track_duration(candidate) < max_duration:
break
playlist.append(candidate)
start = end
start_vec = end_vec
playlist.append(end)
return playlist
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('mp3tovec', type=str, help='MP3ToVecs file (full path)')
parser.add_argument('--inputs', type=str, help='Text file with list of songs')
parser.add_argument('output', type=str, help='Output MP3 filename')
parser.add_argument('n', type=int, help='Number of songs to add between input songs')
parser.add_argument('--noise', type=float, help='Degree of randomness (0-1)')
parser.add_argument('--epsilon', type=float, help='Epsilon distance (default: 0.001)')
args = parser.parse_args()
mp3tovec_filename = args.mp3tovec
tracks_filename = args.inputs
mix_filename = args.output
n = args.n
mp3tovec = pickle.load(open(mp3tovec_filename, 'rb'))
mp3_vec_j_norms = {} # precalculate norms for mp3_vec_j
for track_j in mp3tovec:
mp3_vec_j_norms[track_j] = np.linalg.norm(mp3tovec[track_j])
noise = 0
if args.noise is not None:
noise = args.noise
epsilon_distance = args.epsilon
if epsilon_distance == None:
epsilon_distance = 0.001 # should be small, but not too small
input_tracks = []
if tracks_filename is not None:
with open(tracks_filename, 'rt') as file:
for track in file:
input_tracks.append(track.replace('\n',''))
else:
user_input = input('Search keywords: ')
while True:
if user_input == '':
break
tracks = sorted([mp3 for mp3 in mp3tovec if all(word in mp3.lower() for word in user_input.lower().split())])
for i, track in enumerate(tracks):
print(f'{i+1}. {track}')
while True:
user_input = input('Input track number to add, ENTER to finish, or search keywords: ')
if user_input == '':
break
if user_input.isdigit() and len(tracks) > 0:
if int(user_input)-1 >= len(tracks):
continue
input_tracks.append(tracks[int(user_input)-1])
print(f'Added {tracks[int(user_input)-1]} to playlist')
else:
break
print()
total_duration = 0
if len(input_tracks) == 0:
tracks = [mp3 for mp3 in mp3tovec]
input_tracks.append(tracks[random.randint(0, len(tracks))])
if len(input_tracks) > 1:
playlist = join_the_dots(input_tracks, n=n, noise=noise)
else:
playlist = make_playlist(input_tracks, size=n, lookback=3, noise=noise)
tracks = []
for i, track in enumerate(playlist):
tracks.append('-i')
tracks.append(track)
total_duration += get_track_duration(track)
if n == 0 and i == 0 or n != 0 and i % (n+1) == 0:
print(f'{i+1}.* {track}')
else:
print(f'{i+1}. {track}')
print(f'Total duration = {total_duration//60//60:.0f}:{total_duration//60%60:02.0f}:{total_duration%60:02.0f}s')
print('')
print(f'Creating mix {mix_filename}')
pipe = sp.Popen(['ffmpeg',
'-y', # replace if exists
'-i', 'static/meta_data.txt'] + # use this meta data
tracks + # append playlist tracks
['-filter_complex', f'loudnorm=I=-14,concat=n={len(playlist)}:v=0:a=1[out]', # normalize and concatenate
'-map', '[out]', # final output
mix_filename], # output file
stdin=sp.PIPE,stdout=sp.PIPE, stderr=sp.PIPE)
out, err = pipe.communicate()
if pipe.returncode == 1 and err is not None:
print (err)