CharacterGen

Overview

An AI-powered character card generator and editor with intelligent context handling and cascading regeneration capabilities.

CharacterGen is a Python-based application for creating and editing v2 character cards. It uses a prompt system that combines base prompts with user input and contextual information to generate character attributes in a customizable order.

Features

• Character Management:
  • Load and save character cards in JSON or PNG format
  • Edit generated outputs
  • Single field regeneration
  • Cascading regeneration for dependent fields
• Ease Of Use:
  • Field focus, a way to expand a field or multiple that you are working on to a much larger text area.
  • Ability to generate/save multiple first messages with one as the main first message and the rest going into alternate greetings.
  • Append generated message examples easily.
• Base Prompt System:
  • Save and load different prompt sets
  • Conditional content based on user input
  • Context-aware field references
• Intelligent Context Generation: Uses a tag-based system for precise context placement
• Flexible Generation Order: Customizable field generation sequence with dependency handling

Generation Tab

Base Prompt Tab

Planned Features

(in progress)

• Editing of the other fields making up a v2 card such as creator, version, tags etc.
• adding more conditional tags such as {{if_no_input}}
• Configuration tab to pass some simple generation perameters like max tokens, as well as be a place to include other configuration options.
• Saving the user prompts into the card to be repopulated when loading a card, this could also be done with the basePrompts used for generating that card.
• AICharED made by Zoltan, AVA, and Neptune has alot of nice features that I like but may never get around to adding them. I would enjoy implementing their better UI formatting, and maybe QoL features such as:
  • a tag cheat sheet with quick copy
  • info on hover for each fields function
  • token counter with breakdown

Installation

Prerequisites

• Python 3.x

Setup

1. Clone the repository:

   git clone https://github.com/yourusername/CharacterGen.git
   cd CharacterGen

2. Configure the API settings in config.yaml. For atleast Ooba and KoboldCCP don't forget the /v1/chat/completions, the default should work for most people:

   API_URL: "http://127.0.0.1:5000/v1/chat/completions"
   API_KEY: "YOUR_API_KEY"

3. Run the application:
   • Use start.bat
     • This creats a virtual env, and gets prerequisites

Usage

Base Prompts Tab

The Base Prompts tab is where you configure the generation templates for each character field. Current default prompt does not include generating personality field.

Available Tags

• Basic Field Tags:

  • {{input}}: User input insertion
  • {{name}}: Character name
  • {{description}}: Character description
  • {{scenario}}: Scenario information
  • {{first_mes}}: First message
  • {{mes_example}}: Message examples
  • {{personality}}: Personality traits

• Ignored Tags(Passed over and not replaced:

  • {{char}}: Ignored since they are used within silly tavern to replace Character names
  • {{user}}: Ignored since they are used within silly tavern to replace User names

• Conditional Input Tags:

  {{if_input}}
  Content only included when user provides input
  {{/if_input}}
  

Generation Order

• Set the order by numbering fields (1-6)
• Only reference tags from fields that come earlier in the generation order
• Fields without an order number are skipped during generation

Generation Tab

Character Management

• Load/save characters using the top controls
• Character files are stored in the data/characters folder
• Save completed characters using the "Save Character" button
• To save as PNG make sure to load a picture first, otherwise a default is used

Field Generation

• Name Field: Toggle between direct input and generated name
• Other Fields: Input is optional, used as context in generation
• Generation Controls:
  • "Generate All": Sequential generation of all fields
  • 🔄: Regenerate single field
  • 🔄+: Cascading regeneration (updates dependent fields)

Tips

• Save different base prompt sets for different character types
• Use conditional tags to handle optional input gracefully
• Set generation order to build context progressively
• Use cascading regeneration to maintain consistency