Chargen V2

🚀 CharGen v2

CharGen v2 is a powerful model designed to assist users in crafting characters for role - playing. It generates character descriptions step - by - step in a dialogue format, reducing repetition and enabling partial re - rolls of character fields.

🚀 Quick Start

Live version

You can access the live version of CharGen v2 at https://chargen.kubes-lab.com.

✨ Features

• Step - by - step Generation: Unlike its predecessor v1, CharGen v2 generates character fields one at a time, which helps reduce repetition and allows for partial re - rolls.
• Custom Dataset: It is trained on a custom non - synthetic dataset that is manually curated.
• Multiple Field Support: Supports generating various character fields such as description, scenario, personality, first message, and dialogue examples based on the Tavern Character Card V2 spec.

📦 Installation

No installation steps are provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

Here is an example of the character generation process:

Advanced Usage

The model uses a dialogue style for generating characters field - by - field. Here are the prompts for different fields:

System prompt:

You are an expert in creating interesting roleplay characters.

Description:

Here is a brief overview of a character. Expand it into a detailed description. Include details about character's personality, their outfit and figure. Mention their age and gender, if applicable.

Scenario:

Write an interesting and engaging scenario for roleplay between Maria and User.

Personality:

Write several personal qualities that characterize Maria.

First message:

Write the initial message in this roleplay that would introduce User to Maria.

Dialogue examples:

Write a few example exchanges between User and Maria in chat format. Separate each exchange with a <START> tag.

<|im_start|>system
You are an expert in creating interesting roleplay characters.<|im_end|>
<|im_start|>user
Here is a brief overview of a character. Expand it into a detailed description. Include details about character's personality, their outfit and figure. Mention their age and gender, if applicable.
Fuckana is a friendly and talkative catgirl that has enormous breasts. Her voice is very quiet and low. She is very smart and will teach User JavaScript - even if they resist it. She has orange fur on her tail and wears full diving suit.<|im_end|>
<|im_start|>assistant
Fuckana is a female catgirl. She has two cat ears, feline - like eyes with slit pupils, and long purple hair. Her breasts are absolutely enormous, even for her body's large size. She stands 6 feet tall in height. Her skin is entirely covered by a full diving suit, except her head. The rest of the catgirl is completely black, except for an orange stripe on her tail. Her voice is very quiet and low - she speaks like a sleepy woman. But despite this, she talks excessively and never stops talking once she begins.<|im_end|>
<|im_start|>user
Write an interesting and engaging scenario for roleplay between Fuckana and User.<|im_end|>
<|im_start|>assistant
You're relaxing on the beach when Fuckana appears from the water and insists on teaching you javascript.<|im_end|>
<|im_start|>user
  ...

📚 Documentation

Quantized versions

• GGUF

Prompts

The model uses a dialogue style to generate characters field - by - field. It supports the following fields based on the Tavern Character Card V2 spec:

• Description
• Scenario
• Personality
• First message
• Dialogue examples

The model does not use {{user}} and {{char}} placeholders. Instead, address the user as "User" and the character by their name.

Methodology

CharGen was created to assist the author (Kubernetes Bad) in writing characters. The following processes were involved in making CharGen:

Data gathering

CharGen was trained on data from Chub, Venus, and JanitorAI character cards. Initial scraping was done between August and September 2023, and an update scrape was performed in November, adding about a third more cards. The data was stored in a MySQL database. After scraping, the total dataset had just over 140k records.

Pre - filtering

A series of SQL scripts were used to discard "bad cards" that couldn't be used for training or would require too much effort to fix. These included broken cards, Spanish - language cards, non - plaintext cards, and cards with very low or exceedingly large total token counts. After deduplication, the total set was cut down to 16k cards.

Minor tweaks

A cursory manual review was performed to eliminate non - plaintext cards that weren't caught by automatic pre - filtering. Name adjustment was also done, and non - English cards were removed.

Data grading

A custom tool was used to grade the cards as "good", "bad", or "to fix". The card selection criteria were defined to ensure the quality of the dataset. It took about 800 hours for one person to grade the cards.

Grammar pipeline

A combination pipeline of a Coedit model and a Llama2 - based model was used for grammar correction. The efficacy of the models was calculated based on several metrics such as deletion and addition lengths and the number of spaces in deleted and added segments.

🔧 Technical Details

Model Base

The base model used for CharGen v2 is mistralai/Mistral - 7B - v0.1.

Dataset

Warning

⚠️ Important Note

This model was trained on some NSFW content, so it may produce NSFW characters.

📄 License

CharGen v2 is licensed under the Apache - 2.0 license.