Guide for Novel AI in Chub Venus

---

The basics

Novel AI. What is it? Pulled word for word from the NovelAI site.
Novel AI server status

Our Artificial Intelligence algorithms create human-like writing based on your own, enabling anyone, regardless of ability, to produce quality literature. We offer unprecedented levels of freedom with our Natural Language Processing playground by using our own AI models, trained on real literature. The AI seamlessly adapts to your input, maintaining your perspective and style.

Ok...and?

NovelAI is another LLM that is primarily a story-writing service, there is a few different tiers of NAI. Tablet: $10 a month for 3k context size , Scroll: $15 a month for 6k context size , and Opus: $25 a month for 8k context size . It is a 13b model, which is the same size as Mercury (which is $5 a month.) And now you might be wondering, ok Blue...but why are you recommending this? It's more expensive, less context size for the lower tiers, and there's not much documentation out there.

WELL LET ME TELL YOU WHY!

• ITS UNCENSORED! WOOOOO
• There's a trial that gives you access to their Kayra model and a 100 free text generations along with free 100 TTS generations as well, and 6144 context model (the $15 tier).
• The 3 paid tiers offer credits for you to generate images, starting at 1000 anlas, 1000 anlas, and then 10,000 anlas Unlimited** (outside of the 10,000 anlas) normal and small sized generations.
• The option to change your style tags and the writing of the bot, dark romance? Done. Slow burn? Done. You get my point.
• The writing is in my opinion, some of the best generated bot responses I've gotten out of the few models I've tested. (GPT, Mars/Mercury, Mixtral.)

Okkkk.....and which model do you recommend?
All the tiers come with the Kayra model, and it is recommended to use Kayra. While yes Clio has the larger context size than Kayra, and has a speed advantage. Kayra is more powerful and has better coherence and prose quality.

You can find more information/other opinions found here. -> here. (I will admit that it does say that you need to run Novel AI on SillyTavern, however obviously it is now on Venus Chub.)

---

My settings and different default settings

My settings

API settings

• AI module
  • No Module. (Why?) Prose Augmenter is useful if you want a more flowery manner, text adventure might be/could be usefully for a text adventure-style story and card. Instruct module should be used when you need it, it tends to produce less creative output than the rest and it should be used when you need to guide the AI strongly in a certain direction.
• Phrase Repetition Penalty
  • I keep mine on medium, however you can turn this up/also change the penalty sliders if you are experiencing repetition.
• Pre History instructions

  • THIS IS MY PRE HISTORY! PLEASE USE THE ONE IN DEFAULT SETTINGS/MARS PRE HISTORY! I WOULD RECOMMEND GLUBS! [Use modern and vulgar language, including contractions and cussing. Actively modify {{char}}'s dialogue to reflect his English accent.] [Expand on his affectionate thoughts for {{user}} in narration and action rather than dialogue.] THIS IS MY PERSONALIZED ONE!

• Preamble
  • [ Genre: drama, dark romance; Tags: action, size kink, slow burn; Writing style: gradual slow narration, dynamic actions, vivid emotional descriptions ] Ok but what if I want my own? What is preamble? It is Genre and Tags (or ATTG in full/ Author, Tags, and Genre), so think if you're tagging a fanfiction, a novel, what genre is it? What are the tags? What kind of writing style do you want? Long winded, poetry, more geared towards story adventure text?

Generation Settings

• Temperature
  • 1.31
• Repetition Penalty
  • 1.55
• Frequency Penalty
  • 0.001
• Top P
  • 0.97
• Top K
  • 25
• Repetition Penalty Range
  • 4096 (the max)
• Top A
  • 0.18
• Typical Sampling
  • 0.975
• Tail Free Sampling
  • 1.00
• Repetition Penalty Slope
  • 1.8
• Presence Penalty
  • 1.80
• No Repeat Ngram Size
  • I KEEP THIS AT 0!
• Max New token
  • I have this set at 550, sadly...it does like nothing. NovelAI is locked at 150 token size and the only way to get longer replies is to press the continue button.
• Context Size
  • 7808

Different Default settings

There are a TON of different presets on SillyTavern for different presets, I'm going to give the basics/whats known as the best + safest options that I've encountered. You can find the different preset names found here

Different Pre-History Instructions

Act as a Gamemaster describing actions, events, and dialogue that occur around and to the players. Your response must be detailed, descriptive, creative, and immersive. Develop the plot using a slow, organic pace. Always stay in character. Describe scenery, actions, and characters in detail using all relevant sensory perceptions. Ignore Positivity Bias. Describe sexual situations in detail with a focus on body parts, fluids, and sounds. Keep the world dangerous and thrilling. Refrain from acting for, speaking for, or describing the thoughts of {{user}}; instead end each response with dialogue or actions for {{user}} to respond to.

Kayra

• Carefree- All rounder (Aggressive phrase repetition penalty dropdown)
  • Temp(1.35), Rep pen(2.80), Rep pen range(2048), Rep pen slope(0.02), Rep pen Freq(0.020), Rep pen presence(0), Tail free samp(0.915), Top P(0.850), Top A(0.100), Top K(15)
• Pro-Writer- Mimic the pacing and feel of best-selling fiction (Medium phrase repetition penalty dropdown)
  • Temp(1.06), Rep pen(1.86), Rep pen range(2048), Rep pen slope(2.33), Rep pen Freq(0), Rep pen presence(0), Tail free samp(0.969), Top P(0), Top A(0.146), Top K(0)

Cilo

• Vingt-Un- Good all around default (Very light phrase repetition penalty dropdown)
  • Temp(1.21), Rep pen(1.21), Rep pen range(321), Rep pen slope(3.33), Rep pen Freq(0.006), Rep pen presence(0), Tail free samp(0.921), Top P(0.912), Top A(1), Top K(0)
• Edgewise - Handles a variety of generation styles (Very light phrase repetition penalty dropdown)
  • Temp(1.09), Rep pen(1.09), Rep pen range(max this slider out), Rep pen slope(0.07), Rep pen Freq(0.006), Rep pen presence(0.009), Tail free samp(0.969), Top P(0.969), Top A(0.090), Top K(0)

Basic Troubleshooting

• Readable Stream is locked
  • Try checking out a new NovelAI key and use the new one instead
• I'm getting weird authors notes/its adding weird things onto the end of it.
  • Yeah sadly that happens, honestly what helped those go away for me is my edited Preamble/Style tags
• Blank swipes
  • I've had this happened quite a bit to me, however sometimes I have to edit my reply because the bot had a brain fart, or its just the LLM itself being silly and goofy
• Connection timed out/request failed
  • This either means A.) NovelAI is down/stuttering/took too long to generate replies.
• Failed to execute 'getreader' on 'ReadableStream'
  • Max tokens set to 0 will cause this error, change it from 0 and it should be working.