RPG Character Generator

The primary scene of the asset. Players fill out character traits, an AI generates a backstory and portrait, and then they can chat with the character in-context.

How It Works

  1. The player fills out character attributes (name, race, class, universe, alignment, level, context) or leaves them random
  2. A prompt is constructed and sent to the selected LLM backend (LlamaCPP / Ollama / OpenAI)
  3. Simultaneously, a portrait prompt is sent to Stable Diffusion (Automatic1111) or OpenAI image generation
  4. The character appears with their backstory and portrait
  5. The player can then chat with the character using full roleplay context

Scene Setup

The primary configuration game object is VelesioAI Integration. It contains:

  • LLM connection settings — Configure your LlamaCPP / Ollama / OpenAI server
  • Automatic1111 connection — Configure your Stable Diffusion server
  • Scene settings — UI state management and panel references
ℹ️
Local LLM: You can also use a local LLM through the disabled LLM object — enable it and uncheck "API" in the main configuration object.

LLM Configuration

Within the VelesioAI Integration object, find the LLMCharacter script. Configure:

  • Host — Your LlamaCPP server address
  • API Key — Authentication key for the server

Stable Diffusion Configuration

Below the LLM config you'll find the Automatic1111 configuration:

  • Hostname — Your SD server address (may be the same as LLM if using Velesio AIServer)
  • Models — Press "List Models" to load available models from the server
⚠️
Important: You must press "List Models" after configuring your SD server to populate the model dropdown. Models are loaded at runtime in the active scene.

Portrait Generation

The character portrait uses the SD to Image script found on the portrait game object:

  • Model — Which SD model to use for generation
  • Prompt / Negative Prompt — Built automatically from character attributes at runtime
  • Editor Generation — You can test image generation directly from the Editor inspector

Dropdown Options

All dropdown options in the scene are designed to work with the recommended model stack. They are trigger words that the portrait LoRA expects. Visit the LoRA's CivitAI page for the full keyword list.