‘Tis the season for some deep carols

Since I am hearing xmas music all over the place, why not have my little char-rnn session music model complete some well-known tunes of the season!

I prime the system with the beginning of “Jingle Bells” (which is actually in its training data in the same key):

M: 2/4
L: 1/8
K: Gmaj
|: BB B2 | BB B2 | Bd GA | B4

and there it goes away:

M: 2/4
L: 1/8
K: Gmaj
|: BB B2 | BB B2 | Bd GA | B4 |
BB B2 | Gc c2 | BB B2 | A2 G2 :|
|: DF G2 | EG E2 | DG G2 | AB c2 |
dB G2 | FG A2 | FGAB | d2 c2 :|

What about “O Holy Night”? I prime it:

M:6/8
L:1/8
K:D
F3 F2 F|A3- A2 A|B2 B G2 B|d6

and it produces something plaintive, but not even close:

M:6/8
L:1/8
K:D
F3 F2 F|A3- A2 A|B2 B G2 B|d6-|d2 A B2 A|d6 |
c2 d d2 A|d3 d2 B|A2 A A2 B|A3- A2 d|
e3 e2 f|g3 g2 f|e2 c d2 c|d2 c3 d|
e e2 c2 A|A3 G2 A|B2 A B2 c|d2 c d2 e|
f3 f2 d|d3 f3|e3 e2 f|e4 d2|f3 f2 d|e6|B3 c dc|B3 B2 A|
B3 B2 A|F2 E F3|B6-|B4 f3|
A3 A2 c|e2 c c3|B3 B2 c|d3 d3|

“Silent night”?

M:6/8
L:1/8
K:C
\"C\"G>A G E3|G>A G E2z|\"G\"d2 d B3|\"C\"c2 c G2z|

Not really capturing the spirit, except maybe the new age spirit:

M:6/8
L:1/8
K:C
"C"G>A G E3|G>A G E2z|"G"d2 d B3|"C"c2 c G2z| "Em"g>d e d2e|"Am"c>BA
A2B|"C"c>BA "Em"B>GE|"C"e2"C"d "C"cFG "G"G>AB|"Am"A>G A
"Em"B2G|"C"c2A "G"G>EG|"D"A>GA "Bm"F3:|

I prime the system with the beginning of one of my favorite tunes, “We Three Kings Of Orient Are”:

M:6/8
L:1/8
K:Em
\"Em\"B2A G2E|\"B7\"FGF \"Em\"E3|\"Em\"B2A G2E|\"B7\"FGF \"Em\"E3|

and out it creates something really on its way to fantastic!

M:6/8
L:1/8
K:Em
"Em"B2A G2E|"B7"FGF "Em"E3|"Em"B2A G2E|"B7"FGF "Em"E3|
"Em"B2A G2F|"C"EDE "Bm"F2B|"C/A"E2E "Em"D2B,|"Am"A,3 "F#1"E2||
|: B/2c/2|"Em"d2B d2B|dgf "C"e2d|"Am"BAG "D7"A3|"G"B2d "C"e3|
"G"GAB "D"d2c|"G"BBB "D"AGF|"Am"E2F "G"GAB|"Am"ABA "G"G2D:|

With a few changes, that becomes the basis for my new deep learning-assisted music composition, “We three layers o’ hidd’n units are” (probably a new hit for the season!):

Advertisements

3 thoughts on “‘Tis the season for some deep carols

  1. Pingback: Paper of the Day (Po’D): The Turing Test and the Evaluation of Generative Music Systems Edition | High Noon GMT

  2. Pingback: “Lisl’s Stis”: Recurrent Neural Networks for Folk Music Generation | High Noon GMT

  3. Pingback: Taking A Christmas Carol Toward the Dodecaphonic by Derp Learning | High Noon GMT

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s