INPUT

Quick character description - Users specify some character traits to be associated with a character. The program then will pick words associated with that trait to be used with that character. Words could be pithy sayings, actions, or even adjectives.

https://github.com/tra38/Skynet (Disclaimer: I wrote this as part of my original attempt at narrative generation, but the results was panned. Handling character descriptions was one of the highlights...in the sense readers never complained about my characters. At the same time, people never seemed to noticed or cared about the characterization at all, so it's possible that this approach may have been unnecessary and a less heavyweight solution should be preferred.)

Setting - When writing the story templates the computer will end up using to generate the story, set aside some spaces for describing the setting. You could attempt to just plug in just the name of the setting ("EwanG's Glorious Emporium") into those blanks, or maybe you could have the computer look up information in a database such as Wikipedia or WordNet to give it more flavor. This approach was used in the computer-generated novel "Around the World In X Wikipedia Articles" that uses Wikipedia to describe its settings (https://github.com/dariusk/NaNoGenMo-2015/issues/142) and in a series of simulations fantasy novels that uses WordNet to describe its settings (https://github.com/dariusk/NaNoGenMo-2015/issues/40).

Alternatively, you could generate descriptive and evocative phrases associated with your setting name. This can add more flavor to your story. (This was done in the computer-generated novel "Our Arrival": https://github.com/dariusk/NaNoGenMo-2015/issues/25)

General Outline Of The Story - The 'story compiler' approach, invented by Chris Pressey, starts off with a "null story" (characters meet each other and then characters leaves). The compute starts filling in the outline by adding in plots to this "null story", but each plot has a wildcard that can allow for plots to be nested within plots. After it finishes adding plots to the story, it then goes through each scene of the story and then "flesh" it out.

The Story Compiler approach was used to generate "A Time of Destiny". Here's an issue that gives you links to the theory, the code, and the actual novel: https://github.com/dariusk/NaNoGenMo-2015/issues/11

Now, you want to specify the plot, not let the computer randomly decide, so you may want to skip the first few steps, while instead focusing on having the program handle your unique array of scenes. This is going to be the hardest part of your program (writing up the different plots and all the individual scenes within each plot). And you don't want to write up too many plots, because then it will be seen as you writing your stories, not necessarily the computer. You could have plots reuse different scenes to reduce your workload.

OUTPUT Output of Couple Hundred Pages - Every year, programmers all around the world attempt to participate in National Novel Generation Month, where the goal is to generate 50,000 words. This is both a very easy problem (set the program up in a loop until you get over 50,000 words) and a very hard problem (once people see the pattern within the generated story, they will get bored pretty fast and stop reading...or even skimming...how do you prevent readers from getting bored when reading the novel?). Still, if you can build a program that can handle one of your arbitrary input, then you can build a program to handle hundreds of arbitrary inputs. Whether you want to read all of the resulting outputs however is a different question...but then you're planning on 'lightly' editing the output, right? Maybe it might work.

Every link that I gave you was an entry in the latest NaNoGenMo competition (https://github.com/dariusk/NaNoGenMo-2015), with the exception of the Skynet link that I gave you. Skynet, however, was intended to be be used in an entry in that competition.

====

So yeah, what you say is doable, and can be done. If you want to take this as a challenge to produce your own lovely program to generate your 85 wonderful stories, go right to it. (Or just find somebody's novel generator and just modify that to produce your lovely 85 stories).

But don't limit yourself to conventional understanding of stories. For example, one of my actual entries to NaNoGenMo involve randomly shuffling the paragraphs of an article, thereby generating a brand new article (https://github.com/dariusk/NaNoGenMo-2015/issues/180). The result was rather surprisingly effective, and I ended up using this similar approach in generating a blog post (https://gist.github.com/tra38/8a6bf3743cd89687151c) and a short story (https://gist.github.com/tra38/02b03745e7da37789ed2). Experimentation can lead to new insights, especially about creativity.