The files are nearly 6GB and too large for Github, so they have been hosted as a zip file at:
https://archive.org/details/anime-according-to-ai-zip
I started with Scrivener, then moved to Vellum since it had eBook formatting. For the Amazon KDP and IngramSpark print versions, I rebuilt the book in Affinity Publisher. (fuck Adobe)
It came out beautiful. Everything is aligned to a baseline grid, the font choices and weights are exquisite, and the readability is impeccable.
There are two things I'm unsatisfied with:
1 | I couldn't transfer over the hyperlinks from the eBook. On my 8GB M1 Mac mini, it crawls to a halt when I use the Vellum file because there are thousands of embedded links. Just trying to extract the links would take forever waiting for each click to load, let alone embedding them into the PDF one by one. If you are researching anime, I highly recommend you open up the EPUB while listening to the audiobook or reading the print version.
2 | Minor nitpick, the typography is too large in the printed book. I did extensive research to make sure everything was legible, and definitely succeeded in that area, but I probably could have reduced the page count 25-50%.
If you are interested in my process, I kept all of my backup files.
Since y'all are unlikely to have the same software, I've gone back and exported each one into a rich text .rtf file.
Some of them are out of order. The create and modify dates weren't super reliable.
The Introduction changed a ton by the end. Then I rewrote the entire thing from scratch in a few hours over a pot of coffee in a stream of consciousness after recording the audio for the entire audiobook outside at a park. (a church provided the coffee)
The essays constantly changed, then had a mega change when given to Claude 2.0 near the finish line.
It'll be interesting to feed it all into a future AI that can organize and highlight all of the differences in a succinct beautiful chart or spreadsheet.
In the meantime, hopefully some of you will find it interesting or useful as is.
You are welcome to train a local AI on my voice, but I call dibs on SaaS (Software as a Service) platforms like ElevenLabs.
elevenlabs.io/?from=partnerrivera8951
ID: QEj0heL4nQHjaGrihlr0
Use it for whatever you’d like. I won’t sue you. It would be good manners to credit me though.
Reach out to me if you are unsure about your use-case and want signed permission:
I recorded it outside at a park over the course of a month.
DaVinci Resolve did an amazing job removing the noise, but some blue jays were quite obnoxious in their screaming and poked through every once in a while.
I took great care in cleaning the audio meticulously by hand. I removed all breaths and made the timing of the performances emaculate. I even manually removed harsh “s” and similar sounds, which was painfully slow, until I discovered “Sonible smart:deess”.
It took 5 months to edit the 18-hour finished product. Which was particularly annoying considering I was (and currently am) homeless.
While I will never do something that crazy again, it was still a fun learning experience. My ability to feel out ideal timings of silence has been sharpened considerably. I also got really comfortable using Audacity.
It was also quite magical to see first hand how 10 crappy takes can be edited together to make me sound flawless. Even a single take can have the timing of words scootched around to great effect.
DEDICATION
This work is dedicated to the creative generative intelligence that orchestrates order from chaos, and to the world-wide public domain for the benefit of all. I surrender all current and future copyright interests completely and irrevocably under the Creative Commons CC0 license.
“ANIME: ACCORDING TO AI” is free for anyone to copy, modify, publish, use, compile, sell, or distribute, for any purpose, commercial or otherwise, and by any means. You don’t need my permission.
See:
https://creativecommons.org/publicdomain/zero/1.0/
for the full legal details.
Although this work is freely available, it makes references to copyrighted material for purposes of commentary and education under the principle of fair use. Use of third-party copyrighted material is subject to the copyright restrictions of the original owner.
This work is provided "as is", without any warranties (that means direct promises) of any kind, either express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and non-infringement. The author (that’s me) shall not be liable for any claim, damages, or other liability, arising from, or related to, the work or the use or other dealings in the work.
Freely download all versions of this work from:
- stevencasteel.com
- archive.org/details/@stevencasteel
- github.com/stevencasteel