1
00:00:12,024 --> 00:00:17,840
>> Hello. Welcome back to Curlyboi Theatre, 
everyone. I hope you all had a lovely lunch  

2
00:00:17,840 --> 00:00:27,040
or time zone appropriate break. We have 
an exciting afternoon of more fun talks  

3
00:00:28,880 --> 00:00:46,320
ahead of us. So, first up, we have Tsz Tsz Kiu 
Pang sorry. Being a musician, Pang has composition  

4
00:00:48,720 --> 00:01:01,360
in his work. His most recent endeavor is 
using the open software, Abjad, to create  

5
00:01:01,360 --> 00:01:15,680
complexly notated musically. He is working to 
maintain software for interesting musicians. 

6
00:01:18,640 --> 00:01:27,760
TSZ KIU PANG: Thanks, Betsy. Hi. Sorry. Yeah, so,  

7
00:01:27,760 --> 00:01:32,640
today I'm going to talk about how to compose 
algorithmic music with Python and Abjad.  

8
00:01:35,200 --> 00:01:41,280
I would like to acknowledge the traditional 
owner of the land which I'm currently on,  

9
00:01:42,080 --> 00:01:47,280
so the [Indiscernible] people and also the 
[Indiscernible] of the [Indiscernible] nation.  

10
00:01:48,480 --> 00:01:53,760
I would like to acknowledge that they 
are the traditional owners of the land,  

11
00:01:53,760 --> 00:01:59,600
which I'm on and I acknowledge the 
elders, past, present and emerging. 

12
00:02:04,160 --> 00:02:10,640
Yeah. Cool. Sorry. Sorry. Who am I? I'm Tsz. 
And I have [Indiscernible] in my first name.  

13
00:02:10,640 --> 00:02:18,880
So, I work as a DSP software engineer and when 
I'm not working, I usually compose music and  

14
00:02:18,880 --> 00:02:26,080
I perform music sometimes, as well. These are 
my socials if you want to look for me online. 

15
00:02:26,720 --> 00:02:33,600
All right. So, in today's talk, I would first, you 
know, give some sort of background information to  

16
00:02:33,600 --> 00:02:44,000
algorithmic composition, just a couple examples 
and then we'll look into LilyPond, which is a  

17
00:02:44,880 --> 00:02:50,960
music engraving software, then we'll be looking 
at Abjad just a little bit more in detail,  

18
00:02:52,000 --> 00:02:58,240
including how to extend Abjad, as well. 
Then, I will talk about how to structure  

19
00:02:58,960 --> 00:03:04,640
your composition as a Python package, at least 
like the way that I do it, the way that I have  

20
00:03:04,640 --> 00:03:11,200
been doing it. And, yeah, that's it.
All right. Cool. So, the best way  

21
00:03:11,200 --> 00:03:16,960
to introduce algorithmic composition 
is to give a couple of examples. So,  

22
00:03:16,960 --> 00:03:24,880
algorithmic composition is actually a thing that 
has been there for a pretty long time, you know,  

23
00:03:24,880 --> 00:03:32,240
much longer than computer was here. So, even 
in time, back in the 18th century, I think,  

24
00:03:33,040 --> 00:03:39,440
there was a pretty famous dice game. 
So, the dice game go like this...so,  

25
00:03:41,120 --> 00:03:47,200
you can probably see the figures on the slide. 
So, the top picture, like we have kind of a magic  

26
00:03:47,200 --> 00:04:00,320
square looking squares and music that Mozart has 
composed. So, the performer or another composer,  

27
00:04:00,320 --> 00:04:06,000
the procedure would be is something like this. 
They would have to roll two dice and use the  

28
00:04:06,000 --> 00:04:15,200
sum to look up a measure number. So, think of 
measure number as an index to a Python list. 

29
00:04:15,200 --> 00:04:20,320
So, basically, you know, using the sum to look 
up a measure number and then they would repeat  

30
00:04:20,320 --> 00:04:29,360
the process for however many times it's required 
and finally concatenating a new piece of music.  

31
00:04:30,240 --> 00:04:40,160
So, the general idea is kind of like to scramble 
the music by, you know, random chance. So,  

32
00:04:40,160 --> 00:04:46,320
you can see how, like, you know, there is a system 
and different procedure to this composition or  

33
00:04:46,320 --> 00:04:52,160
to this dice game. And it is very much doable by 
computer, but obviously, like, people didn't do it  

34
00:04:52,160 --> 00:04:59,760
by computer. I imagine I don't really know, but I 
imagine they were doing it by paper and pen. So,  

35
00:05:00,400 --> 00:05:12,080
yeah. So that was one example. In the early 20th 
century, we have serialism, which is, you know,  

36
00:05:12,080 --> 00:05:18,320
a new school of floats in music composition 
and one of the really prominent techniques,  

37
00:05:18,320 --> 00:05:23,520
in serialism, is the [Indiscernible] technique 
and I'm really simplifying things here. 

38
00:05:23,520 --> 00:05:31,520
So, we have a turn row. So, this is kind of, 
like, the basic idea of the technique, which is  

39
00:05:31,520 --> 00:05:37,440
based on a turn row, which is a series of nine 
repetitive pitches. Apologies for those who do  

40
00:05:37,440 --> 00:05:46,640
not read music. You can see the 12 different notes 
on the slide there. So, they are nonrepetitive in  

41
00:05:47,200 --> 00:05:52,800
a series and the composer can apply operations 
on a turn row to generate music materials  

42
00:05:53,440 --> 00:06:03,920
and transposition, inflection and retrograde. 
And the composers would apply these operation,  

43
00:06:03,920 --> 00:06:10,240
any of these operations to generate 
music materials for their composition. 

44
00:06:11,840 --> 00:06:18,720
So, we're looking to transposition first, we have 
the original turn row. I am labeling each page by  

45
00:06:18,720 --> 00:06:25,440
a unique page class number, which is ranging 
from zero to 11 because we're dealing with  

46
00:06:25,440 --> 00:06:33,680
[Indiscernible] technique. In western classical 
music, in western music, in general, it is very  

47
00:06:33,680 --> 00:06:44,320
much based on, you know, [Indiscernible] 
like in an octave, zero to 11. So, yeah. 

48
00:06:45,600 --> 00:06:53,200
With transposition, we're pretty much 
preserving the row and moving it up or down  

49
00:06:55,600 --> 00:07:05,440
and moving the turn row up by a turn, so, by a 
little bit up. And, the operation that we apply  

50
00:07:06,640 --> 00:07:13,840
to the original turn row can actually be done by 
just one line of Python code, so you can like,  

51
00:07:13,840 --> 00:07:22,160
you know, on my scene here, we have pitch, 
plus two, 4 pitch in original turn row. So,  

52
00:07:22,160 --> 00:07:28,480
that's, you know, that's the Python code that 
we have to perform to transpose. Pretty simple.  

53
00:07:29,040 --> 00:07:35,600
It is very much doable by a computer. Of course, 
a lot of composer have been doing it by hand,  

54
00:07:35,600 --> 00:07:40,720
which is fine, as well. But, you know, I just 
would like to point out, well, you know, there are  

55
00:07:40,720 --> 00:07:48,320
actually a lot of techniques in music composition 
that can be done by a computer algorithmically. 

56
00:07:49,600 --> 00:07:55,440
All right. So, next up, we have 
inversion. We're presenting a contour,  

57
00:07:56,400 --> 00:08:02,640
but we are flipping it upside down. 
So, again, this is very much doable  

58
00:08:02,640 --> 00:08:06,400
by just one line of Python code, you know, 
just a list comprehension kind of thing.  

59
00:08:07,680 --> 00:08:17,600
And we have five speech, 5 is a magic number in 
here. Of course, if you want to, you know, change  

60
00:08:17,600 --> 00:08:24,000
5 into 6 or 7, like, that is perfectly fine, 
as well. It will just be, you know, transposed,  

61
00:08:24,000 --> 00:08:31,440
like, up or down. So, like, moving up or down. And 
then we the inverted turn row below. So, you know,  

62
00:08:31,440 --> 00:08:38,560
that's how we generate the inverted turn row.
Next up, we have retrograde. So at this time,  

63
00:08:38,560 --> 00:08:45,200
we're preserving a contour. So instead of having 
it upside down, this time, we're having it,  

64
00:08:45,200 --> 00:08:52,160
you know, so we're flipping it sideways and 
again, just one line of Python code and we are  

65
00:08:52,160 --> 00:08:59,360
using the reversed method in Python to reverse the 
lists and the rest, you know, plus two is just to  

66
00:09:00,480 --> 00:09:06,480
move it up by a whole tone, basically. 
Then we have the reverse turn row. So,  

67
00:09:07,680 --> 00:09:11,120
you can see, like, all these 
techniques, although they were invented,  

68
00:09:11,120 --> 00:09:16,960
like, in the early 20th century, but a 
lot of those, like, can actually been  

69
00:09:16,960 --> 00:09:21,840
done by computer. Like, you know, using 
today's technology, especially in Python. 

70
00:09:22,960 --> 00:09:29,920
All right. So, yeah. As I've mentioned, those 
were just a couple of examples of algorithmic  

71
00:09:29,920 --> 00:09:36,800
composition. There are a lot of other techniques, 
as well, so, there are some software which there  

72
00:09:36,800 --> 00:09:44,080
are also some software which help composes 
to compose music. With notation software,  

73
00:09:46,400 --> 00:09:58,240
we have, you know, Sibelius. They're rely good at 
doing their job, which is to notate music. We have  

74
00:10:00,080 --> 00:10:14,400
OpenMusic and Bach. They can use the music to 
generate music material kind of algorithmically,  

75
00:10:16,160 --> 00:10:22,480
but to notate the actual music, like the composer 
would often have to resort to using other  

76
00:10:23,200 --> 00:10:27,280
softwares and kind of similarity 
for notation software, as well,  

77
00:10:28,000 --> 00:10:32,960
a lot of the software doesn't actually 
provide algorithmically composition capability  

78
00:10:33,520 --> 00:10:39,920
to the composer so they would have to resort to 
using, you know, other software or composing by  

79
00:10:39,920 --> 00:10:46,640
hand first to get the idea of what they want 
in their [Indiscernible] and use the software. 

80
00:10:47,760 --> 00:10:52,800
So, you know, that's why Abjad kind 
of becomes really handy, because it  

81
00:10:52,800 --> 00:10:58,160
kind of bridges the gap between the notation 
software and the composition software. So,  

82
00:10:58,880 --> 00:11:05,680
pretty much like everything can be done within 
Abjad. So, both composition and notation.  

83
00:11:05,680 --> 00:11:13,520
So, to get a better understanding of where 
Abjad sits in this kind of software landscape,  

84
00:11:13,520 --> 00:11:21,040
we would have to look at LilyPond first. 
So, LilyPond is a music engraving program.  

85
00:11:21,040 --> 00:11:26,160
So, it is not in terms of the interface, 
it's not too different from Lapec.  

86
00:11:28,960 --> 00:11:36,240
The architecture is obviously very different. 
Here's a block diagram. So, it takes a plain text  

87
00:11:36,960 --> 00:11:56,160
file, a .ly file as the input and it 
can format files. So, like, you know,  

88
00:11:56,160 --> 00:12:05,040
the ly file looks like this. You can see, c4 
and then we have four notes here. That's kind of  

89
00:12:05,760 --> 00:12:11,360
the idea of LilyPond. So, just taking 
plain text, generating music notation. 

90
00:12:12,960 --> 00:12:19,280
All right. So, I thought I would, you 
know, give you a really brief look at, like  

91
00:12:19,280 --> 00:12:26,480
of what a LilyPond file would look like. So, this 
is an example from the LilyPond documentation.  

92
00:12:27,360 --> 00:12:37,600
So, again, like, we have notes here. It notes 
there, just kind of like a circle, like, you know,  

93
00:12:38,160 --> 00:12:47,440
[Indiscernible] and, you know, those are musical 
notes. And, yeah, so it corresponds. We have d4,  

94
00:12:48,320 --> 00:12:54,080
ac and we can also see the number, which 
indicates the duration of each note.  

95
00:12:55,920 --> 00:13:02,800
Then, you can also see, like, there are features. 
These vertical lines, which are, you know,  

96
00:13:02,800 --> 00:13:10,400
corresponds to the line here and we also have some 
key symmetries in here, as well. All those things  

97
00:13:10,400 --> 00:13:16,960
you would have to specify in the LilyPond file and 
it can generate all these features in the score. 

98
00:13:18,400 --> 00:13:28,400
This is another example. You can see, 
like, you know, even with this thing here,  

99
00:13:28,400 --> 00:13:36,400
again, you can specify, in the LilyPond, and that 
will generate yeah [Indiscernible] in the score. 

100
00:13:37,520 --> 00:13:43,920
All right. And you can also do some, you know, 
a little bit more complicated rhythmics in the  

101
00:13:43,920 --> 00:13:53,280
music. You can use tuplet to group three notes 
inside the space of two. Three divided by two  

102
00:13:53,280 --> 00:13:58,640
and so on. This is just like couple of examples 
of, like, what a LilyPond file would look like. 

103
00:13:59,920 --> 00:14:05,760
All right. So, you know, those examples, they 
are pretty simple. You know, they are not  

104
00:14:05,760 --> 00:14:12,480
too many notes. Not too complicated rhythmic 
stuff going on and the file relatively small,  

105
00:14:12,480 --> 00:14:19,440
just a few lines. But you can probably imagine, 
like, if I'm composing a really big piece, the  

106
00:14:19,440 --> 00:14:27,520
file would be pretty large so this is kind of the 
next slide. So, this file is just really large.  

107
00:14:28,720 --> 00:14:33,200
You know, it's so many lines and apologies for the 
font. It's just really small. I'm trying to fit  

108
00:14:33,200 --> 00:14:41,040
in as much as possible on just one slide. And that 
was only about one tenth of the file that was used  

109
00:14:41,040 --> 00:14:48,560
to create this. So, this is my composition. And 
this and this. So, like, not too much music here.  

110
00:14:48,560 --> 00:14:55,280
I believe there are only 21 bars of music in 
here. And, like, it takes a really large file  

111
00:14:56,000 --> 00:15:04,400
a really large .ly file to generate all these. 
So, like, if I'm kind of, like, engraving music,  

112
00:15:04,400 --> 00:15:11,760
that's kind of fine. I know which note knows goes 
where exactly and I know how to take those things.  

113
00:15:11,760 --> 00:15:18,000
I have to do [Indiscernible] type the notes 
into the .ly file and that'll compose for me. 

114
00:15:21,680 --> 00:15:30,320
When it comes to composing, I often have to go 
back and change things in my score and if I'm  

115
00:15:30,320 --> 00:15:37,360
working with such a big file, a big, you know, .ly 
file, and I would have to look for, okay, where do  

116
00:15:37,360 --> 00:15:43,200
I want to change my score? Which note do I have 
to change? It's just like, kind of complicated  

117
00:15:43,200 --> 00:15:50,080
when I'm dealing with the music, note by note. 
So, here is where Abjad comes in pretty handy.  

118
00:15:51,760 --> 00:15:58,000
So, Abjad is a Python Package that helps compose 
complex pieces of music notation in iterative  

119
00:15:58,000 --> 00:16:06,160
and incremental ways and Abjad extends the Python 
language. So, as a composer, I can use Abjad and  

120
00:16:06,160 --> 00:16:13,920
utilize the power of the Python language and also 
other Python packages to assist my composition,  

121
00:16:13,920 --> 00:16:21,440
as well. And finally, Abjad wraps the LilyPond 
notation software, which is demonstrated by  

122
00:16:21,440 --> 00:16:30,240
this diagram here. So, we can create a table an 
instance of Abjad voice, just a Python class,  

123
00:16:30,960 --> 00:16:38,240
so, Abjad and voice and we have a music notation. 
It looks like LilyPond code, but it's not really.  

124
00:16:39,040 --> 00:16:46,160
And Abjad can understand that. It interprets that 
and it can spit out LilyPond code when you call  

125
00:16:48,560 --> 00:16:56,800
"Abjad.LilyPond." So you can call LilyPond to 
compile the code for you and that will generate a  

126
00:16:56,800 --> 00:17:06,800
music notation and even better, Abjad actually has 
the capability to call LilyPond within itself, so,  

127
00:17:07,840 --> 00:17:14,720
all you have to do is to create a table 
an instance of Abjad.voice and that  

128
00:17:14,720 --> 00:17:19,040
will call LilyPond instead 
of Abjad and generate a PDF  

129
00:17:19,040 --> 00:17:25,120
as the output. You don't have to look at the 
LilyPond code at all. Bypassing the process. 

130
00:17:26,880 --> 00:17:32,240
All right. So, to get, you know, a little 
bit more taste of, like, how to use Abjad,  

131
00:17:32,240 --> 00:17:38,560
I guess we'll start with a pretty simple example. 
So, we'll start with just four notes. So,  

132
00:17:38,560 --> 00:17:44,000
on my slides, here, we have of course, if we 
want to use Abjad, we have to import Abjad.  

133
00:17:44,000 --> 00:17:56,640
So, I have a list of pitches, from 0 to 11. I can 
actually extend it to negative number and beyond  

134
00:17:56,640 --> 00:18:02,160
11, as well, but I'm not going to go into 
[Indiscernible]. So, what I'm doing here is to  

135
00:18:02,160 --> 00:18:09,200
create a table an instance of Abjad.duration. 
So, 1/16, that would be a 16th note.  

136
00:18:10,320 --> 00:18:17,040
And for those of you who don't understand music, 
that's fine, as well. So, 1/16th is just basically  

137
00:18:18,640 --> 00:18:25,200
a musical note with a pretty short 
duration so 1/16th of a bar, I think?  

138
00:18:27,120 --> 00:18:32,640
Then we can use Abjad.leafmaker and provide 
it with the pitches and the durations and  

139
00:18:32,640 --> 00:18:40,800
that will generate the musical notes for us.
So, the output of this piece of code kind of  

140
00:18:40,800 --> 00:18:45,040
looks like this. You know, just, like, four 
notes. So, you know, pretty simple, right?  

141
00:18:45,680 --> 00:18:53,040
One almost suspects, well, why do I have to do 
it in Abjad when I can just do it in LilyPond  

142
00:18:53,040 --> 00:19:01,920
with probably fewer lines of code? But here's the 
magic of Abjad. So, you can utilize the Python  

143
00:19:01,920 --> 00:19:08,560
programming language, for example, I'm using some 
list operation. Just like extending my pitches  

144
00:19:08,560 --> 00:19:16,880
list by using a list comprehension kind of thing, 
so I am extending my existing pitches using a full  

145
00:19:16,880 --> 00:19:26,320
loop. This creates a list of [Indiscernible] 16th, 
I think and then I'm doing exactly the same thing.  

146
00:19:27,440 --> 00:19:41,760
Abjad.duration, 1/16 and using the leaf maker. 
Leafs mean the musical notes sorry, the black  

147
00:19:41,760 --> 00:19:49,680
dotted notes. So, the result kind of looks like 
this. This is just me using a full loop in Python.  

148
00:19:50,320 --> 00:19:57,440
So, yeah. You know, you can only see how I can 
extend this technique even a little bit more. 

149
00:19:58,960 --> 00:20:07,040
So, we can even retrograde the Melody by the 
list. After extending the list, I'm extending  

150
00:20:07,040 --> 00:20:17,920
the list and used reversed method. So, here I am 
utilizing one of the techniques, but, you know,  

151
00:20:17,920 --> 00:20:25,920
nothing stopped me from retrograding the Melody. 
So, the result kind of looks like this. So, wow,  

152
00:20:25,920 --> 00:20:30,960
like, we can really see the benefit of using 
Abjad because, like, if I do it in LilyPond, like,  

153
00:20:30,960 --> 00:20:35,920
it would probably take me, like, quite a while to 
figure out, like, how to retrograde everything. 

154
00:20:38,560 --> 00:20:47,840
And we can also play around a bit with rhythm. 
So, specifying, like, 1/16th, I am using a list  

155
00:20:47,840 --> 00:20:58,560
of three different durations, 2/16, 4/16. You 
might ask, well, okay, what happens now? Because,  

156
00:20:58,560 --> 00:21:03,440
like, the length of the duration of the list and 
the length of the pitches list are different so  

157
00:21:03,440 --> 00:21:13,680
I believe the pitches is of length 32, I think, 
and durations is only of length 3. So, huh.  

158
00:21:13,680 --> 00:21:20,720
[Indiscernible] not work but, you know, it will 
actually still work. So, Abjad kind of you know,  

159
00:21:20,720 --> 00:21:27,520
it will cycle through the list and Evan though 
the two lists are of different length, Abjad  

160
00:21:28,720 --> 00:21:32,640
will cycle through the list so the 
result kind of looks like this.  

161
00:21:33,840 --> 00:21:36,560
And now we are doing some, like, 
complicated rhythmic stuff. 

162
00:21:37,440 --> 00:21:43,680
And for those of you who reads music, you probably 
noticed that, oh, there's something wrong with the  

163
00:21:43,680 --> 00:21:54,880
music. If you don't read music, that's fine, as 
well. In this bar line and so, like, there's quite  

164
00:21:54,880 --> 00:22:00,720
a bit of space between this bar line and this 
note over here. So, that's something wrong with  

165
00:22:00,720 --> 00:22:08,160
the music because I am playing too much with 
the rhythm so I would have to fix this problem  

166
00:22:08,160 --> 00:22:16,640
and this is actually pretty easy to fix in Abjad.
So, I can, you know, specify the meter apologies,  

167
00:22:16,640 --> 00:22:22,480
it's musical time for those of you who 
don't read music. But, yeah, in essence,  

168
00:22:23,680 --> 00:22:28,400
I'm just, like, trying to demonstrate, all 
right, there are, like, a lot of useful  

169
00:22:28,400 --> 00:22:38,720
methods and classes in Abjad, here, I am 
using Abjad. You can use, this is kind of  

170
00:22:38,720 --> 00:22:50,000
like almost like really Pythonic kind of syntax.
So, yeah. And then I'm just using the meter class,  

171
00:22:50,000 --> 00:22:55,680
class method, rewrite meter and that will rewrite 
the music for me, according to this meter.  

172
00:22:55,680 --> 00:23:01,600
So what I'm trying to do is fix the duration 
problem which occurred in the previous slides.  

173
00:23:03,440 --> 00:23:11,360
So, yeah. Then we have fixed the duration problem 
and we have a proper piece of notated music and,  

174
00:23:11,360 --> 00:23:16,960
you know, I have been doing this, step by step, 
and hopefully it's not too complicated but you  

175
00:23:16,960 --> 00:23:23,840
can already see how Abjad provides a really 
convenient interface for composer to compose  

176
00:23:23,840 --> 00:23:33,760
music, step by step. I can also add other stuff 
to the music as well. I'm going to add accents.  

177
00:23:39,120 --> 00:23:51,760
I'm using a full loop. It provides some iterate 
functions for us to iterate over the music. So,  

178
00:23:52,960 --> 00:23:59,280
a lot of things are really object oriented. 
I'm treating every note as an object.  

179
00:24:00,080 --> 00:24:07,760
Here, I am iterating over the music and please do 
not worry too much about the logical ties, it's  

180
00:24:09,280 --> 00:24:15,280
more like an [Indiscernible] in there, but 
essentially, I'm I am iterating over the  

181
00:24:15,280 --> 00:24:22,960
music and adding accent for every three notes, 
if you want, or every three [Indiscernible].  

182
00:24:24,320 --> 00:24:30,480
All right. So, I'm using the attach function 
to attach something to the music. So, you can  

183
00:24:30,480 --> 00:24:38,720
probably see, like, I'm adding accent. By accent, 
I mean this triangular looking thing to the music.  

184
00:24:39,520 --> 00:24:46,160
This is really iterative. So, you know, I am doing 
things, like, you know, step by step and I'm,  

185
00:24:46,160 --> 00:24:51,520
like, complicating the music. You know, 
just by a few lines of code every time. 

186
00:24:53,360 --> 00:25:00,560
All right. So, as you can see, the core object 
library provides many useful classes and methods,  

187
00:25:01,680 --> 00:25:08,560
but they are not really enough to cover 
algorithmic composition and techniques. So,  

188
00:25:08,560 --> 00:25:14,800
current extension includes object, which is the 
[Indiscernible] extension and the Abjad makers.  

189
00:25:14,800 --> 00:25:20,960
So I'm not going to talk about it because I don't 
know much about it. I'm going to talk about Nauert  

190
00:25:20,960 --> 00:25:35,680
because I am using it. It is the extension. It 
is based on [Indiscernible]. And it quantizes  

191
00:25:35,680 --> 00:25:42,720
time based events into music notation. It looks 
something like this. The composer would have to  

192
00:25:42,720 --> 00:25:51,040
provide a list of duration and pitches and that 
will be the input to the quantizer. It is very  

193
00:25:51,040 --> 00:26:02,000
much configurable and generate music notation.
So, to use the package, it would be something  

194
00:26:02,000 --> 00:26:06,160
like this. So, pretty much similar to the 
block diagram, which I've just shown. So  

195
00:26:06,160 --> 00:26:15,040
creating a sequence, which is just a sequence 
that the package can understand, by providing a  

196
00:26:15,040 --> 00:26:23,840
list of durations and pitches and calling it with 
the SD input and that will do something like this.  

197
00:26:24,480 --> 00:26:32,880
You will be like, wow, and I really should have 
mentioned the durations are in millisecond. So,  

198
00:26:33,520 --> 00:26:38,720
yeah. One doesn't even have to understand that 
much about music notation and they can, you know,  

199
00:26:40,400 --> 00:26:42,720
write something that is as complicated as this.  

200
00:26:44,480 --> 00:26:52,240
So, this is pretty useful, especially for, 
you know, a top down approach for composition. 

201
00:26:52,240 --> 00:26:59,040
All right. And of course, we can utilize other 
Python package. So, as you can see, I'm using  

202
00:26:59,040 --> 00:27:07,360
NumPy here. I'm defining a function using NumPy 
to generate just a list of exponential distributed  

203
00:27:08,720 --> 00:27:14,720
durations and using NumPy again to 
select a you know, just a list of random  

204
00:27:15,520 --> 00:27:20,240
pitches for me. And finally, using the 
quantizer to quantize the events for us. 

205
00:27:22,080 --> 00:27:28,480
Now we can create music in a rather abstract way. 
By specifying the music duration or the total  

206
00:27:28,480 --> 00:27:36,720
duration of that piece and the average duration 
of each note, we've, you know, in seconds, then,  

207
00:27:36,720 --> 00:27:43,360
you know, finally, providing with a pitch set so 
just a list of pitches and we can have something  

208
00:27:43,360 --> 00:27:50,160
like this and we can also increase the density 
by decreasing the average duration of each note  

209
00:27:51,840 --> 00:27:57,520
and now we have something like this, even more 
complicated and now we are torturing the musician.  

210
00:28:00,880 --> 00:28:07,040
By changing one number, we can have so 
many different music generated for us. So,  

211
00:28:08,320 --> 00:28:13,040
yeah. I've shown you, you know, how to use 
the Abjad and how to use the extension,  

212
00:28:13,040 --> 00:28:21,120
as well. But when it comes to composing 
a really large piece, we can do it in,  

213
00:28:21,120 --> 00:28:25,600
you know, in the Pythonic way, as well. 
We can structure it as a Python package,  

214
00:28:25,600 --> 00:28:36,560
which is something that is a lot of us are doing. 
So you can see, like, we have README, just like a  

215
00:28:37,440 --> 00:28:46,080
regular repository. As you can see, we have info 
and set up the [Indiscernible] which indicates we  

216
00:28:46,080 --> 00:28:53,200
can pypi and store this, if we want to.
So, I'm going into the 47 40_directory.  

217
00:29:01,840 --> 00:29:14,000
I can install them if I want to. The score will be 
built in the builds directory and where I define,  

218
00:29:15,200 --> 00:29:24,640
you know, each [Indiscernible] of my music and 
finally, we can actually py test to see everything  

219
00:29:24,640 --> 00:29:36,080
is working fine. This is my definition.py. So, you 
can see, because I am treating this as a Python  

220
00:29:36,080 --> 00:29:43,520
package, so I can actually import 47 from within 
the directory, which is not part of the Python  

221
00:29:43,520 --> 00:29:50,640
package, but because I'm installing other parts 
of the 47, as a package, so I can import 47 here  

222
00:29:51,680 --> 00:29:57,280
and I can use, you know, some of the modules 
from within the [Indiscernible] directory. 

223
00:30:03,200 --> 00:30:08,960
So, yeah, that was more, or less, about, 
all I can talk about in a half an hour.  

224
00:30:10,080 --> 00:30:18,800
If you want to learn more about Abjad, here is 
the Abjad website and the quantizer documentation.  

225
00:30:18,800 --> 00:30:25,840
There are also other music examples, as well, 
in these few links. A lot of them, they have  

226
00:30:25,840 --> 00:30:33,280
the link to the repository, as well. So, if you 
want, you can have a look at their code. So,  

227
00:30:33,280 --> 00:30:39,200
yeah. That's it. Thank you very 
much for for your time and, yeah,  

228
00:30:39,200 --> 00:30:49,360
I hope you will have a great day after this.
>> Thank you so much Tsz, for that really,  

229
00:30:49,360 --> 00:30:56,080
really lovely talk. I was saying, in the chat, 
that I don't have a whole lot of music experience,  

230
00:30:56,080 --> 00:31:00,720
but I was able to follow that, thanks to your 
lovely explanations and it was really interesting. 

231
00:31:02,640 --> 00:31:05,920
We don't have any time for 
questions, unfortunately.  

232
00:31:06,720 --> 00:31:12,080
Tsz, do you have some time to step 
into the text chat to answer because  

233
00:31:12,080 --> 00:31:17,520
there were a couple of questions entered in?
TSZ KIU PANG: Yeah. Definitely. I will jump  

234
00:31:17,520 --> 00:31:22,320
into the text chat, just in a moment.
>> The Curlyboi Theatre chat,  

235
00:31:22,320 --> 00:31:27,440
in Venueless, I will copy the questions 
over there, everyone. Thank you so much,  

236
00:31:27,440 --> 00:31:33,600
Tsz, for a lovely talk. I know there were a lot of 
other music y and not music y people in the chat  

237
00:31:33,600 --> 00:31:44,065
enjoying it, as well. So, thank you a lot.
TSZ KIU PANG: Thank you very much, Betsy.