Skip to content

SciFi Mythos: Far, the Faithful

October 21, 2010

The cavern city, must be confessed by
the faithful. Now, from the necks, and fell
the king. Thou, what we could ensure,

till the slaves were known; there would not secure.
The Lord said little.
For the Devil stood, down from the centuries. Freedom.

From the wings, the Lord that looked. The deport,
when they could see what the sword cometh. The mad.
Before that stood abyss.

So far, the heaven, who endured.
The tallest peaks.
The walls.
Fear the night.
Far, the faithful.
Superiority of the Lord. Stood.
A leader. Stood on.

The sun,
The Lord thou, from light.
Paradise: the sea.

October 20-21, 2010; stochastic beam search using phonemic evaluation on a bigram word model. Texts: selections from HP Lovecraft, the King James Bible, Gibbon’s Decline and Fall of the Roman Empire, Milton’s Paradise Lost, Dante’s Inferno. Generator: ePoGeeS.

Made with Audacity and Cepstral. Loop: Iron Maiden, “Number of the Beast.”

a couple days ago I checked out the first couple seconds of this video:

and I was like: yeah, grab a guitar and do some poetry!!! but I was like: man, I don’t know how to play guitar… but then I was like: thanks to the magic of technology I can SAMPLE guitars!!1! It’s what Flash would do! I had a hard time building a loop until I found a guitar-only version of Number of the Beast… hey… it’s like some supernatural force WANTED me to write this poem!!!

So I used a variation of an approach I’d used earlier for coaxing narrative out of bigram models. First, I generated a number of lines (14n, where n was 4 or 5) without punctuation, evaluated for alliteraton and the “UH” phoneme. Then I picked out the lines that had words or phrases I liked (around 20). Then I went through the lines, one at a time. For each line, I performed one or more of the following transformations:

  • added punctuation
  • capitalized words
  • removed words from the beginning or end of the line
  • split up the line

and then I listened to the line through a Text-To-Speech program. If it didn’t sound good, I performed the transformations again until it did. I don’t remember there being any lines that I had to give up on, and I decided not to use SSML. As I transformed the lines, I arranged them in an order that suggested a narrative. this is method 6d025542-c970-480c-867c-08be67e38419 (arranging and pruning, adding punctuation to improve TTS).

I still feel a little like adding music is gratuitous, but then I figure:

  1. Fuck it. This is the first (non-song!) poem that I want to bang my head to.
  2. If I don’t do it, people will come up with far lamer approaches.
  3. Dude, it’s in the tradition of ‘respected’ epic poems like Beowulf:

GRENDEL GONGA!!! wha…..??? BLOODEM UN DRANK!!!! yo man, whatever you do, DON’T make Beowulf sit up in offence!

No comments yet

Leave a Reply

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

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

Google photo

You are commenting using your Google 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 )

Connecting to %s

%d bloggers like this: