My AI Process

First, how I used AI image creation in the process of OCaK.

I’m gonna go through this in the process I remember using AI image generators in roughly chronological order. From my memory, I first started using image generators to create visuals of character appearances. A bit ironic but I’ve struggled with describing character appearances if I don’t have a visual. I’ll have an idea of certain aspects, such as eye color or skin color or hair type, but there’s a wide range any of those could fall under. When I say a character has curly hair, what type of curly hair am I trying to describe? When I say a character has blue eyes, do I want to describe them as more of a striking blue or a more muted blue? I personally don’t know until I have a visual for myself.

I’ve tried things like Picrew or other character creation games but the available options were always a bit more limited than I like and I was never able to fully “get” what I was imagining in my head. So eventually I started using NovelAI’s image generator to make the first drafts of the characters. I’d often recreate them with each update to NovelAI’s image generator there but eventually got the first versions of a lot of my characters.

The next thing I ran into was the fact that sites such as Campfire, where I was kinda planning on putting characters and other information, doesn’t allow for AI generated images (understandably!). Additionally, as I developed OCaK further, things weren’t…fitting with the NovelAI images anymore. Mainly clothing and scars on characters and it started to bother me.

That led to the second step in the process I’ve been using. I’ve made basic headshots of the Main/Primary characters already without the usage of AI and even made a small base for myself with different face shapes, eye shapes, ears, and noses (I ended up not using the noses further on but figured I’d mention it!). So using my own art as a reference, I put in the NovelAI image into ChatGPT and asked it to recreate the image in a similar style to my original art. It took a bit but eventually came up with something I liked. I then recreated/redrew the image in Clip Studio. Some elements of the AI image were re-traced, fixing and altering aspects along the way, as well as adding my own art. I refer to this art/images as “AI Redrawn” or “AI Referenced/Assisted” since while I redrew the image myself, AI was the base image.

For any future image that I mark as “AIRedrawnArt” (I’ll describe the tags a bit further), assume I’ve used this process of generating an AI image, likely either with ChatGPT or with NovelAI, then put that image into Clip Studio and redrew/traced/recreated the image from there.

Now from my memory, somewhere in the process of figuring out the character appearances and getting a more solid idea of what each story’s plot will roughly be, I wanted to come up with covers for each of the stories. Since I was still brainstorming, I used ChatGPT to generate first draft book covers of the stories that would end up being the Main Series and Secondary Series. I used these as placeholders for a while before purchasing a Canva subscription, at which point I then recreated the covers, changing them to fit a sort of “style” I wanted between the covers.

I wasn’t 100% happy with the center image/icons I created in Canva so I did take those center images, put them into ChatGPT to create cartoon style versions that I then put into Procreate, tracing over different elements of different images to take what I liked from each. I then replaced those center images in the Canva version and that’s how I’ve gotten the current version of the covers I now use.

The last major thing I used AI image generation for is some concept images for OCaK. Primarily, the clothing for the cultures in OCaK. Since I didn’t want any culture in OCaK to be a “replacement” culture that is a downfall of a lot of fantasy world building, I tried to pull inspiration (specifically for the clothing) from IRL cultures of similar geographical location as are in OCaK. I had a few vague ideas for the cultures that would be “in focus” for the stories but wanted to get at least a starting version of the other cultures so then I can (hopefully) avoid any plot holes or struggling to create those in the future.

So my process was prompting ChatGPT with a variety of information (IRL cultural clothing inspirations, biomes the culture inhabits, population in each biome, temperature, and any other vague ideas I had originally) then generated concept images for 4 “basic” types of clothing: lower-class every day wear, upper-class every day wear, lower-class formal wear, and upper-class formal wear. Then, similar to what I did for the character art, I recreated these AI generated images, fixing and altering as I saw fit until I was pretty happy with the result.

Tags and Watermarking: I know how it can get tiring on the internet trying to detect if something is AI or not, especially when it comes to images. I don’t want people to stress too much here so I’ll be using a bit of a system to be as clear as possible.

When possible, I’ll use tags to clearly define images and art. For reference, here are the tags that will be used and their level of AI involvement:

1. #AIGeneratedImage - The image is fully AI generated. There’s been no human alterations made to the image. I’ll also use #ChatGPT and/or #NovelAI alongside this tag to indicate which generator I used.
2. #AIRedrawnArt - This is for images such as the headshots and full body references I’ll be making for my characters. For images/art here, there was an originally AI generated image I then redrew either in Procreate or Clip Studio. The full image will be redrawn by me without the original AI image attached.
3. #OriginalArt - This is for my art with absolutely no AI involvement As well as not using any purchased art bases (since I want that to be a separate tag). This means no referencing AI or generating images.
4. #BaseArt - This is for images where I used purchased bases/lineart. Credits to original base/lineart owner will be provided. No AI will be used in these arts/images.
5. #CopyrightFreeImage - This is for images used that are free to use and/or have a creative license that allows usage without credit. These are primarily from sites such as Usplash.

Additionally, in an effort to avoid people taking any AI generated images from this blog, I’ll be placing watermark on AI images that simply says “AI Generated Image”. This is not in an effort to prevent people from “stealing” any AI images from this blog since 1) I legally have no copyright of the AI generated image and 2) I’ll never claim ownership of the image itself. This is an effort to prevent someone taking that image and attempting to pass it as their own “art” and/or profit from it.

Next, onto how I used AI in the plot, outlining, and worldbuilding of OCaK.

Again, I’m going to try to do this in chronological order, similar as before. OCaK originally started with a few non-related AI chats using Claude within SillyTavern and character cards. The two earliest stories, "The Flower and the Bull" and "The Viper's Poison" both started as non-OCaK character cards. However, as I began to form the idea of these unrelated character cards, as well as thinking about other stories I’d like to write about in the world of OCaK, these character cards/chats weren’t working how I wanted them to.

So, I created my own AI characters cards for SillyTavern, used the World Info feature to provide the worldbuilding I was creating, and started a new chat/RP with those cards. These new OCaK cards and their chats/RPs would help me develop the overarching story, plot points, and allowed me to bounce ideas around (and able to find out what I did and didn’t like for my stories). These AI character cards and chats are what I’d personally consider the very first rough rough rough drafts of their respective stories. Due to the heavy usage of AI in these, I don’t plan to post these anywhere.

Once I got a start to what I think I’d like each story to be, my usage of AI for the outlining and writing process dropped off greatly. For the outline of all the stories, no AI was used beyond me going to the AI chats/RPs to look through plot points I enjoyed/wanted to include. I used Obsidian for the development of OCaK (it’s even what I typed this disclaimer up in first) and I used the canvas feature in Obsidian to outline my stories.

When I get to the first drafts of the stories using the outline, I will not use AI generated text for the vast majority (95% of the written text is original) of the draft. I will never use any AI to write full scenes and put those into this first post-outline draft. The 5% AI I gave is that I do plan on using Grammarly and/or ProWritingAid during my writing process and may or may not use the suggestions either program comes up with. Both Grammarly and ProWritingAid use trained AI models so "technically" their rephrasing/rewriting suggestions could be considered AI generated. Just to be "safe", that's why I put the drafts and published stories may contain up to 5% AI generated text.

Next, I did use AI to develop some worldbuilding aspects of OCaK. Primarily, I used AI as a jumping off point. Example: The final major religion group on the western continent of Vitallis was a struggle for me. I had no idea how I wanted this religion to be, their symbolism, beliefs, nothing. The only ideas I had was 1) the "origin" religion has since died out, 2) there are two sects/denominations of the origin religion existing to modern day, and 2) one of the existing religions was influenced by an outside religion. I basically put that into GPT-4o and after a few regenerations, one idea was about a "Great Weaver" of sorts and I ended up really liking the idea. I jumped from there to form that Religion Group. Could I have eventually gotten to this same point without AI? Probably but it really cut down the time I spent trying to come up with an idea I liked.

That's the main way I use AI for worldbuilding. I also used AI to generally check if an idea is plausible and/or realistic. Mainly to do with timeline, traveling distance/time, and land/ocean sizes. I did my best to double check information the AI provided, as I’m well aware of the hallucinations it can have.

TLDR: I use AI generated text to form overarching plot points/plot of books, very VERY rough drafting, grammar/spell check, and starting points for ideas. No AI generated "clumps" (I.E. paragraphs of text) of any type will be used publicly and everything will be re-worked with my own writing.

"Special" Note/Disclaimer: Campfire

As of writing this, I do plan to host content on Campfire Writing eventually, as well here and on AO3 (writing only). Campfire does not allow any AI generated images or text to be published to their platform and I 100% intend to stay within these guidelines. However, I did clarify with Campfire support a few aspects of their guidelines. Campfire support clarified to me that programs such as Grammarly and ProWritingAid are not considered AI and are not included in this policy. I will still say that since Grammarly and/or ProWritingAid technically still use AI, I will still disclaim my writing may contain upwards of 5% AI generated assistance/writing. They also clarified that my process of redrawing/tracing an AI image in Clip Studio would result in the final non-AI image to be allowed as part of a published work.

Side Note: thanks Campfire support for dealing with all of my specific questions about AI usage on your platform! The person who was answering my questions was a patient saint and explained the policies wonderfully.

Finally: AI Usage in the creation of this website:

I did use ChatGPT to assist me with parts of this website. I have created I think 2 websites before, one was for my final in a web development class in college and the other was a personal website. I remember for certain the website for college was 100% not AI, as it was before ChatGPT was released. The personal website I think was assisted by AI in parts, mainly to figure out why the HTML and CSS weren't doing what I was expecting.

I originally was going to use the personal site was the template for this site but then I came across a few sites on here that had used petrapixel's layout generator. Immediately ditched that idea and turned to that generator for the template of this site. From there, I used AI in a moderate degree to develop the rest of this site. AI was a great TOOL: helping me with JS, figuring out why the f* my header would disappear (love forgetting that header.js and Header.js will path differently, it's my favorite thing ever /s), helping me figure out the best font colors when I diverted from anything other than black and white, etc. However, I don't think I would have been able to create this site the way it is without my previous knowledge of how to make a website completely without AI. I can look at my HTMLs, CSSs, and JSs files and (mostly, other than JS still getting re-introduced there) understand what's going on. I feel like I could have gotten to this point without AI, especially since I've done it before. However, I don't think I would have been able to in as quick of a timeframe.