A fictional oral history

[orbific]

My work in progress is in my github repo

I'm planning to use an LLM to generate a novel-length fictional oral history (similar to World War Z or Chuck Palahniuk's novel Rant). While ChatGPT etc struggle to produce strong literary prose, they seem to be very good at first person accounts - as you'd expect from the number of forum posts they must have read.

As of 2024-11-02 I have a version of my code that produces a novel-length text, divided into a number of chapters. I now need to review this and see how it can be made more cohesive.

[orbific]

I'm working away at this... getting slightly distracted by generating 'The Great Gatsy 2: Gatsby vs Kong'.

[orbific]

I just published a blog post on my experiences so far. I'm enjoying NaNoGenMo but I still have a long way to go!

[orbific]

Here is a sample text of my work so far. The main issues I have with this are:

1. There's not a wide enough range of people being interviewed in this oral history - it's got stuck on the initial set.
2. There's a peculiar blandness to the text. Sometimes interesting moments occur, but there's a flatness to the prose.

Technically I've completed the NaNoGenMo challenge by producing a book-length work, but I want to try making a few tweaks before I finish.

novel.pdf

[orbific]

Here is the finished version of the novel. It comes out at 59,500 words.

malarkey-an-oral-history.pdf

The code is maintained within my github repo

There are a few formatting inconsistencies and other things I would work on if I had more time, but it is a complete, computer-generated work of fiction.