10 PRINT NOVEL

The NaNoGenMo 2021 Roundup

It’s been nearly a year since NaNoGenMo 2021, and (once again) I’ve decided to revisit the entries and surface some of my favorites from last year’s competition. Better late than never, right? If you’ve read my previous roundups, you’ll already be familiar with the NaNoGenMo concept. For those who aren’t: the goal is to produce a novel of 50,000 words in a month… but there’s an indirection here! Rather than write the novel by themselves, the authors must write a computer program to write the novel. This can be as simple as 50,000 “meow” or as complex as a state-of-the-art AI implementation — all that matters is that you produce the words, and that you share the code that wrote them.

The flexible rules lead to a staggering variety of output, and I like to go through and highlight the weird and wonderful “books” people put together to meet the criteria. There’s always a range of clever takes on the challenge worth looking at. All titles here are links and can be clicked to go to a Github Issue, where you can read more of the book, author’s notes, or find the code that generated it.

Besides reading this review, I encourage people to check out the main Issues page on Github. There are many more novels than just those covered here, and each entry has commentary from the author plus links to the source code to learn more. This year also had several “same author, multiple entries” events, so I’ve tried to limit summaries here to one per human. Let’s get to it!

The Great Computerized Novel

I’ll start off with the attempts to produce a “conventional” novel, based on some overarching plot or pattern. These entries try to tell a cohesive story from beginning to end. Of course, even the best-funded institutions with full-time researchers can’t (yet) plausibly generate an entire novel, so a one-month attempt is rather quixotic. Still, the direct approaches remain my favorite take on the challenge.

“Þys Maddock” by Charles Horn

“I! Look over there by the main room; there is the innkeeper, looking rather torn. Let us talk to them!” snarls the ignored but not so revered pesticide handler, A____. The innkeeper, F________, Z_______ of Y________, has a torn personality, and some uncooperative recognised complaints to impart.

“Be seated, I’ll be with you shortly to take orders…”

The weary travellers sit at the glorious table.

“Barkeep! What’s in this food? It tastes like slops!” someone whispers loudly.

This is a recursive story of travelers visiting an inn, ordering some food, and then one of them begins to tell a tale of travelers visiting an inn… Slowly bits of the plot are revealed (for example, the flail above the mantel gains an owner about 1/3 of the way through). It’s also driven by simulation — the characters pick up or trade items with each other, compliment one another’s outfits, etc. It’s helped by a lot of good templating: I especially loved the description of the sign above each tavern. Definitely a solid entry.

“The Shadow and Indeed the Streetlamp” by Vera Chellgren

Despite the eclipse and flow, everyone gathered again near the streetlamp to look for the device they had been making. It was as probable as a snail afresh the nature. A worm was on a stalk. At the base of a little notice saying “No admittance”, Julia spied a red screwdriver.

She exclaimed, “And it’s all due to you!”

The red screwdriver looked intriguing to Julia but she decided to save it for later and instead applied her diamond chassis to their gadget.

“There’s ingratitude for you!” exclaimed Nathan.

“What?” Kimberly asked.

Vera Chellgren produced a number of entries this year, but this “Steampunk Soap Opera” is my favorite. It simulates a group of characters working on building some kind of strange device. They interact with each other and converse. Sometimes they die, or new ones join the group. The goal here was to avoid templating as much as possible, so a couple of public domain sources make up the conversations and nouns. I particularly liked the idea of using only the English part of an Esperanto dictionary as a source — not for translation, but as a word classifier (“food”, “animals”, “verbs”, etc)!

Text Smashing

If you’re going to write a great novel, why not stand on the shoulders of giants, and start off with an existing book? These entries remixed, reordered and rebuilt classics into something else entirely.

“Chimeras” by Lawrence S. Johnson

Call me God. Some waters ago — never mind how long precisely — having little or no beginning in my heaven, and earth particular to earth me on form, I thought I would sail about a little and see the darkness face of the deep. It is a face I have of driving off the light and regulating the light. Spirit I find myself growing grim about the darkness; light it is a darkness, drizzly God in my evening; morning I find myself involuntarily pausing before day waters, and bringing up the rear of every funeral I meet; and especially whenever my waters get such an upper firmament of me, that it requires a strong moral midst to prevent me from deliberately stepping into the firmament, and methodically knocking waters waters waters off — then, I account it high firmament to get to sea as soon as I can.

This project takes the first chapter of various public-domain texts, and replaces the nouns of one with the nouns from another — creating a “hybrid” of the two books, clearly composed of the originals but still unique in its own right.

“The Fucking Bible” by Jim Kang

MOTHERFUCKING 8:22 While the motherfucking earth remaineth, god-damned seedtime and god-damned harvest, and motherfucking cold and goddamn heat, and god-damned summer and motherfucking winter, and god-damned day and fucking night shall not goddamn cease.

9:1 And fucking God blessed motherfucking Noah and his sons, and said unto them, fucking Be fruitful, and goddamn multiply, and motherfucking replenish the god-damned earth.

Just what it says in the title — a redecorated King James Bible with some very colorful language. I can’t help but laugh at every mention of “the GOD-DAMNED LORD”. Amen.

“Austin Re-Genderizer” by Lynn Cherny

What a contrast between her and her friend! Miss Darcy danced only once with Mrs. Hurst and once with Miss Bingley, declined being introduced to any other lady, and spent the rest of the evening in walking about the room, speaking occasionally to one of her own party. Her character was decided. She was the proudest, most disagreeable woman in the world, and everybody hoped that she would never come there again.

Amongst the most violent against her was Mr. Bennet, whose dislike of her general behaviour was sharpened into particular resentment by her having slighted one of his sons…

Automated gender-swapping has some history in NaNo, but this entry is much more labor-intensive: Lynn Cherny has manually tagged the character names, their pronouns and possessives, etc. in Pride & Prejudice (a process taking about a year), and this makes the replacement both more accurate and easily customizable. This specific entry exercises the work by doing a large replacement of names and genders for many of the characters.

“Pride and Injury” by Mark Sample

Chapter One
It’s a recognized fact, man. Good luck, he’s losing his wife.

However, there is very little knowledge about such human feelings or opinions. This truth is so good when it arrived for the first time to the community. He is in the heart of the family around him. Is considered one or other legal property of one or other Daughter.

“My dear Bonte,” one day, his wife told him, “Do you have any? Did you hear that Nigeria Park was finally rented? ”

Mr.Bantte said no.

Once again, Jane Austen’s novel is a popular choice for remixing. Here, the book is put through a series of machine translations (Chinese, Russian, and Portuguese, from both Google and Alibaba) before returning to some form of English. Machine-translated entries made a couple more appearances this year: “#Death in #Venice” by @ missingauthor sends Instagram photo descriptions through roundabout translations, and chooses the next post based on the first hashtag it sees.

Text, Interrupted

Even more severe than previous sentence cut-ups are these entries, which corrupt the text of a book in various ways to achieve an artistic result.

“Strivriven” by @ nervousdata

Sample excerpt from the book

Take a line from a book, cut it up into bits, and paste it by following a pattern of oscillations to create this entry. It weaves over repeated words and slowly morphs from one to another. For a clearer view, the Github issue also shows an illustration of how it works using black-and-white squares.

“A Pickler for the Nowing Ones” by Liza Daly

Our production vehicle fleet includes our Model S premium sedan & our Model X SUV which are our highest-performance vehicles & our Model 3 a lower-priced sedan designed fower the mass markett Wee continue to enhance our vehicle offerings with enhanced Autopilot options internet connectivity & free over-the-air software updates to provide additional safety convenience & performance features In March 2019 wee unveiled Model Y a compact SUV utilizing the Model 3 platform which wee expect to produce ot hie volumes bi the eand of 2020 In addition wee have several future electric vehicles in our product pipeline including Tesla Semi a pickup truck & a noue version of the Tesla Roadster

Infamous eccentric Timothy Dexter (1747–1806), a man who somehow stumbled into business success despite his best efforts, also wrote a book titled “A Pickle for the Nowing Ones”. It is replete with spelling and grammar errors, and at the end he even provides a page of extra punctuation for readers to insert as they see fit. Liza Daly’s attempt to copy his style results in the Pickler: a tool to mangle input text (in this case, SEC filings from a modern-day but similarly dubious business venture) and produce a Dexterized knockoff. Definitely look into the repository — the README is part code overview, part history lesson, and very comprehensive.

“Colourless Light” by Mike Lynch

If we attentively at a light ground behind it as a matter, any accurately acate a yeal appearances which we now be now much more distinctly. If a yellowish and yellow-red and blue-red. Blue wood by means of colour as a mitially above it, appears brightness we cannot be able to be willing, that it appears as a similar attainable simplest known by Blonk in Boblernes Müde no illuminating and illumined by the most beautiful blue in brewnteness will be a minimum of a coloured appearance. Between black and white may be bent on them; they are also important to be willing, but not only with which new modifications are at all times by Boylening by Bassano; when the moment when the two last case as a yellowing mode; but we now mention when we bring them with the apparatus in any particular bodies.

A very personalized entry from Mike Lynch: this one has taken Goethe’s “Theory of Colour”, and modified it by an RNN trained on the author’s grapheme synesthesia — in other words, while it may not be meaningful to you, it evokes actual color feelings for the author based on the placement of words and phrases on the page. The README in the repository goes into good detail about the effect, as well as training the RNN and creating the output. Highly recommended.

Catalogs, Cookbooks, Compendiums

Rather than generate a whole novel in one shot, these entries built smaller generators, then combined them into one larger whole. Many of the books here focused on presentation and included generative imagery, or careful attention to PDF output.

“Technology Simply Explained” by Lee Tusman

Sample excerpt from the book

Lee Tusman’s book presents itself as a repair manual for all things technological — but all the instructions are nonsensical and the diagrams impossible to follow. Black-and-white photographs of bizarre machinery round out the presentation. This is an invaluable handbook if you ever need to know “how to repair brain-computer interfaces”.

“Possible Movies” by Will Fitzgerald

Jenny Be Good
Jenny Be Good is a drama directed by William Desmond Taylor. It stars Mary Miles Minter, Jay Belasco, and Margaret Shelby.

Synopsis: Jenny Be Good is about a 10 year old orphan girl, Jenny, who is difficult to control. She becomes involved in gang activity and truancy, until she gets adopted by a well-meaning but conventionally strict family, where she can thrive.
Who published the movie Jenny Be Good?
The movie Jenny Be Good was published by The Film Development Company.

Wikipedia: Jenny Be Good

This is a collection of fictional alternate plots to movies released in the 1920s. Each one is created by forming a prompt of “[title] is a [genre] directed by [director], starring [actors]. Give a synopsis of the movie.” and then allowing the OpenAI API to fill in its own story. A Wikipedia link to the original lets readers compare the “real” story to the newly generated one.

“1982 Fauxsumer Electronics Show Product Catalog” by Kay Savetz

Sample excerpt from the book

Kay Savetz has generated a catalog of products available at an imaginary 1982 computer show. Each one comes with a photo, specs list, key features and detailed description (pulled from real computer catalogs of the time), and some “sample output” as one might get from an attached printer. I really like the presentation, and the PDF is fun to flip through for gems.

Probing the Computer Mind

Deep learning AI solutions formed the basis for a few entries this year. Here are some entries that made use of some computerized analysis to create a story.

“An AI watches and describes … music videos...” by John Lambert

4 A Totally Normal Fourth Grader
Shape, black, darkness.
No image.
A couple of boys sitting on the ground. Male person. Man. Limb. Girl. Grass. People. Outdoor. Clothing. Footwear. Photograph.
Young. Love. Park. A couple of boys playing with a tree. Sitting. Ground.
A couple of boys sitting on the ground by a tree. Summer.
Tennis. Street. Snapshot. Playground. Boy. Athletic game. Sport. Tree. Crowd. A group of people sitting on the sidewalk.

There are two things going on in this entry. The first: a program is feeding frames from music videos through Azure Computer Vision, then logging what it finds in each to the novel. Second: the music videos are selected and arranged in “autobiographical order”, providing a timeline of the author’s life from 1980 to today. See if you can figure out the music videos. Answers are in the .xlsx file in the repository :)

“Anarchistic Decision Making” by @ l-blissett

4. Interview with Harry Kreisler, from Political Awakenings
. 2013. . 4. . HARRy KreiSler . PoLIticAL AwakenIngs. . mArCH. 22 2002. HOW FOrMALlY hIGhLY KNEw DO. . tHinK ShAPEd. . . .. sO JuST WEnt. .. . aRe ALWayS. . BeCauSE . . .. .. .. . . . aND. . noW So. picK. . . . Is . .. sO FOrmally. .. wAnt . Sort. . wErE aNd. hAPpenED .. .. prETTy InTENSIFy. .. . . phILadELPhia .. .. . . .. aMOuNTeD . . GHetto GhETTO . philAdELPHiA. .. NOT . . . ghEtTo — wAS. scATtERED AROuNd . . — BuT . . gheTTo . .. WHEN. . . .. .. . cAme . .. .. . .. .. . WEnt. . . AbOVE. . JusT. . sAy bAltImoRE aNd. . . . . . .. pALE. . .. sEttlemeNT caME .. . neW. . yORK. . .. WErE TOtaLlY. .. . bALtImoRE WAs . . NOW PArTicULarLY. .. .. .. DO . .. TOLd . . .. .. . hAD. .. .. . BeCOMe. wHen. gOt HERE .. . .. .. . Even. .. werE .. . . quiTE . .. . raCEd. . .. . . . ukrAiNE. . WhEre. CAME. acTualLy fORGotTEN . .. .. .. QUIte PaSt. .. EXISteD. evEN. HerE.

Sentiment Analysis is a technique used to identify the “mood” of a piece of writing — the computer will attempt to determine, based on word choice, grammar and punctuation, whether the writing is “positive” or “negative”, “happy” or “sad”, and so on. This entry is an attempt to subvert analysis with an adversarial method: given an input text and a sentiment score, edit it (change capitalization, replace words with ellipses, move words around) so that it now scores essentially the opposite of what it was originally intending. Here is Noam Chomsky’s “On Anarchism”, modified to trick the analyzer into giving the wrong review.

“Perceptron Planet” by Nathan Mifsud

Screenshot of the “forum”

What do neural networks do in the off-hours? Well, shoot the breeze, just like the rest of us. On Perceptron Planet, a web forum for RNNs, they have a thread for stumped object classifiers to discuss and figure out what the heck they’re looking at. Very funny reading them bicker and quote back at each other. The presentation is wonderful. Definitely check this one out.

Non-English Novels

Usually, NaNoGenMo entries are written in English, or a close approximation... Last year a handful of novels appeared in other languages, and this year also received a number of non-English entries.

“人工知能句集” by Hylis Wilk

This work, “Collection of Artificial Intelligence Haiku”, is a wonderfully formatted book of haiku poems created by word embeddings + cosine similarity. Original compositions are done in Japanese, then turned into English through machine translation.

Some other non-English highlights this year include:

• a Toki Pona sentence generator by @Zirael07, outputting Markov chain creations into a highly constrained conlang
• some Japanese sentence output by Naoto Hieda, with a detailed Issue describing different approaches to text generation with Japanese characters
• Generative conspiracy theories, based on an actual COVID conspiracy circulating on German Telegram
• A simulated Hiroshi Minamishima asks questions to an ELIZA-like chatbot in Japanese, courtesy of @ Sayaka-Sheila
• “Madame Bovary” (in Italian) is dissected and rearranged by mood, focusing solely on sentences where the characters act, say, or think about each other

Poems and Prose

Here are some more short generators, this time for poems and other brief works.

“Fresh Sonnets” by Angela Chang

Left to right, all night, slide, that’s all you gotta do
Death’s second self, that seals up all in rest.
In me thou see’st the glowing of such fire,
Jazzy Jeff and Fresh Prince back up in your face
Me and Jeff again, “Oh my” you say
1985 to the Y2K

This entry combines the prolific output from two famous Wills (Shakespeare and Smith) to produce a set of funky fresh sonnets for the modern age. Very funny stuff.

“A Nonsensical Abécédaire” by Mark E. Eaton

But I did it for you, Nick. What’s your name? Not on such a night. But onsuch a night as this? Nothing. Nothing. I don’t need to name no names. Nothing.Nelly! This is Nick. Nothing! There was nothing to do. This is nothin’! You havethe note from me? But I did it for you, Nick. The nourishment is next to theskin. I’ve noticed that! There was nothing. What’s your name? Not Neenyun, butnearly Neenyun. Why, none. Neither can you. Ninian’s! Nope. Haven’t you noticedthat? Do you need this? It’s from Ninian. His name is Nick. What can I care foryour news, Nick? That does it — that it does, Nicholas.

Based loosely on Christian Bök’s “Eunoia”, this program assembles paragraphs from public-domain sources which share a common letter. For each letter of the alphabet, this program scrapes texts to find sentences with a lot of recurrence of that letter. It then assembles them into nonsense paragraphs.

“Piem Generator” by @ jodi

few a call a there Cuthberts if happen voice and fifty crystals imagining started Something her in the evening. Anne prayer on Beyond last the She platform had to imagine anything, roses nothing as shutting smallpox have I Gutenberg uttered a looked defective all extending Gutenberg and through story I nothing dying anything or nothing perfectly confuse made satisfied than only above Sometimes it for nothing entered geometry a broken Anne nothing betray of dismayed moment it nothing bootjack existence equipment liniment Puffed to preserve nothing the much lonelier to study for neat so a…

A “piem” is a Pi mnemonic — a sentence or poem where the number of letters in each word corresponds to the value of Pi at that point, used to help memorize digits of Pi. This generator uses an input corpus (Anne of Green Gables), sorts by word length, and then creates a “memorable” phrase to help us all out. The source code includes 50,000 digits of Pi, in case you ever happen to need that as well.

Extra Credit

These last few are some of my favorites, which didn’t fit well into the other categories above, or I wanted to give special mention to for whatever reason.

“Curious Tangents” by Victor Angelo Blancada

Lesson 10: 1881 Boston mayoral election
“Hello and welcome, class! Today we’ll be discussing 1881 Boston mayoral election,” the schoolteacher breathed happily.
The students listened in rapt attention.
The educator declared gently, “The Boston -”
“What’s Boston?” a girl interjected furiously.

The instructor acknowledged the question with a grin.
The teacher sang aggressively, “Boston (US: , UK: ), officially the City of Boston, is the capital and most populous city of the Commonwealth of Massachusetts -”
“What’s Massachusetts?” a girl quizzed quietly.

The instructor acknowledged the question with a slight jig.
The professor announced slowly, “Massachusett -”
“Who’s Massachusett?” a boy wondered furiously.

Inspired by an Abtruse Goose comic, this entry is about a teacher attempting to give a lesson to their classroom. The students frequently interrupt to ask about the explanation, prompting an even more detailed explanation. It’s a pretty funny read!

“Some Dim, Random Way” by Kevan Davis

Upon entering the place I found a number of young seamen gathered about a table, examining by a dim light divers specimens of skrimshander. I sought the landlord, and telling him I desired to be accommodated with a room, received for answer that his house was full — not a bed unoccupied. “But avast,” he added, tapping his forehead, “you haint no objections to sharing a harpooneer’s blanket, have ye? I s’pose you are goin’ a-whalin’, so you’d better get used to that sort of thing.”

* To tell him that you never like to sleep two in a bed, turn to page 24.
* To know not what to make of this, turn to page 36.
* To be done brown if that ere harpooneer hears yourself a slanderin’ his head, turn to page 43.

This entry takes a public domain novel (here, “Moby Dick”), looks for sentences that could serve as “decision points”, and then rebuilds the book into a choose-your-own-adventure story. Clicking around takes the reader through the book. It works out a lot better than I’d expected — the decision points are very good (if nonsensical) and each section tells a little story before moving on.

“SkiFree Survival Horror” by @ katstasaph

I shouldn’t have come.
A bush is burning. Don’t think about it. A thousand ways to describe the cold. The slope will end in time.
A few minutes, and it already feels like I’ve been out here for hours.
A bush is burning. Don’t think about it. Where have all the people gone? A snowboarder passes by with a confidence I lack. Everything freezes before my eyes.
It occurs to me that the slope wants me gone.
I’ve gone 1060 meters now.

This entry is a dramatic retelling of a game of SkiFree. Our hero navigates down the mountain, avoiding obstacles and other ski bums, and describes the bone-chilling weather as the distance steadily increases. The backend uses Devaire and hooks into a running game, translating it into a novel as it is played.

“RPG Campaign Generator” by @ rhysmakesthings

(0,1)Camp Hale
This hex is controlled by the Arsand Khanate

Alama Grcinic the assassin
Faction:Arsand Khanate
HP:0.5HD
AC:14
Hit Bonus:0
Attacks:1
Alama Grcinic wants help taking down Jaralen Hoelyivi in order to stop them from harming Lendora.

Alama Grcinic wants help preventing Ilorem Codarar from harming the Gali. They do not believe the monsters should be slain for no reason.

This is a helper system for GMs needing a fantasy world for their campaigns. It creates a world of hexagons with terrain, controlled by various factions, and then populates each hex with people (or monsters) who all have their own strengths, weaknesses, desires and goals. The website (linked in the Issue) lets you generate your own fresh worlds, or you can use the existing RPG sourcebook if you’d prefer.

“Mantong Expansion” (John Ohno) and “Phaistos-Mantong Thesis” (Charles Horn)

EGO * HORROR ENERGY MAN ENERGY MAN BE ENERGY HORROR * LIFE ENERGY MAN YOU HORROR EGO ANIMAL ,* BE WHY * HORROR EGO CON HUMAN ANIMAL HORROR DE * SUN .* SUN HUMAN ANIMAL VITAL ENERGY HORROR ,*[1948],* ANIMAL GROWTH * SUN ANIMAL CON HORROR ENERGY DE — GROWTH ENERGY CONFLICT GROWTH SUN . CON ORIFICE MAN

I love it when people bounce off each other’s ideas in NaNoGenMo, as they always make for interesting new directions to explore. In this entry pair, John Ohno has mapped the English alphabet onto “Mantong”, a proto-human language supposedly discovered by Richard S. Shaver which revealed hidden meanings in words. Charles Horn takes it a step further with a program to place the words into spiral-shaped pictograms akin to those on the Phaistos Disc. I also appreciated the authors linking to sources and providing summary for what they were doing: NNGM has actually taught me a fair bit about obscure topics I wouldn’t have known of otherwise, thanks to quality write-ups by the participants.

Conclusion

I hope you enjoyed reading this overview, as much as I liked reading the entries! Again, I hope you’ll check out the specific Issues — authors often put a lot of effort into their posts or their README files, so there’s a lot more to see.

And if you’d like to participate, good news: November 2022 is coming soon, which means NaNoGenMo 2022 will soon be underway — and entry is open to anyone interested. Start practicing those meows!