Skip to content

Settings Guide

This guide explains the global settings that control the underlying AI models. These settings are found in the main Settings menu and generally apply to all your adventures, unlike the In-Game Options Menu which has settings specific to each save file.

Text Generation

These settings control the AI that generates the story text.

  • Provider: Choose between running the AI locally on your computer (Local) or using a custom third-party service (Custom).
  • Context Window Size: Determines how much of the previous story the AI remembers when generating new text. A larger window means better coherence but may be slower.
  • Custom Base URL: If you select the Custom provider, enter the API endpoint URL here.
  • Custom Endpoint: The specific API format for your custom provider (e.g., TextCompletions, Ollama).
  • Temperature: Controls the creativity of the AI. Lower values make the text more predictable, while higher values make it more random.
  • Max Tokens: The maximum length of the text the AI can generate at once.
  • Invert Translation Order (Experimental): This setting changes how the AI processes translations. By default, your input is translated to English, the AI generates a response in English, and then the response is translated back to your selected language. When this option is enabled, the AI generates a response directly in your selected language, and the response is then translated to English for internal processing. This can make the narrator sound more natural and avoid translation inaccuracies.

Local Text Settings

For a detailed walkthrough on setting up local text generation, see the Local Text Generation Guide.

  • Models Folder: The folder on your computer where the local text generation models are stored.
  • Model: The specific .gguf model file to use for text generation.
  • Keep Model In Memory: Keep the model loaded in your computer's RAM/VRAM for faster generation, but uses more resources.
  • GPU Layers: The number of model layers to process on your GPU. More layers mean faster generation but require more VRAM.
  • Backend: The processing engine to use. CUDA is for NVIDIA GPUs, Vulkan is for AMD/Intel GPUs, and CPU uses your main processor.

Cloud Text Settings

  • Model: Select the text generation model offered by your chosen cloud provider.

Image Generation

These settings control the AI that generates the illustrations.

  • Provider: Choose between Local, ComfyUI, or various cloud services.
  • Remove JPEG Artifacts: A post-processing option to clean up compression artifacts, especially for certain models.

Local Image Settings

For a detailed walkthrough on setting up local image generation, see the Local Image Generation Guide.

  • Models Folder: The folder for the main image generation models (.safetensors, .ckpt, etc.).
  • Model: The specific model file to use.
  • TAESD Path/Model: Settings for a tiny autoencoder that can speed up image decoding.
  • LoRA Path/LoRA: Settings for LoRA (Low-Rank Adaptation) models, which can modify the style of the main model.
  • Cfg Scale: How strictly the AI should follow your visual description.
  • Dimensions: The resolution of the generated image (e.g., 512x512, 1024x1024).
  • Sample Steps: The number of iterations the AI takes to generate an image. More steps can improve quality but take longer.
  • Sampler: The algorithm used to generate the image.
  • Keep Model In Memory: Keeps the image model loaded in RAM/VRAM.
  • Scheduler: The noise scheduler used during image generation.
  • Quantization: The data format for the model's weights.
  • Backend: The processing engine to use (CUDA, Vulkan, etc.).
  • VAE Tiling: A memory-saving option for the VAE (Variational Autoencoder).
  • UNET/CLIP/VAE Paths & Models: Advanced settings for specifying individual components of the image generation pipeline.

ComfyUI Settings

For a detailed walkthrough on setting up ComfyUI, see the Using ComfyUI Guide.

  • Endpoint: The URL of your running ComfyUI instance (e.g., http://127.0.0.1:8188).
  • Workflow: The workflow_api.json file to use for generation.
  • Save to Disk: Whether ComfyUI should save the generated images on the server machine.

Cloud Image Settings

  • Model: Select the image generation model offered by your chosen cloud provider.

Depth Map Generation

These settings control the generation of 3D depth information for the 2D images, creating a parallax effect.

  • Provider: Choose between Replicate (cloud) or Local.

Local Depth Map Settings

  • Models Folder: The folder where depth estimation models are stored.
  • Model: The specific model file to use.
  • Keep Model In Memory: Keeps the depth model loaded in RAM/VRAM.

Speech Generation

These settings control the text-to-speech (TTS) for the narrator.

  • Provider: Choose between Cloud, LocalVITS, or LocalJETS.

LocalVITS/JETS Settings

  • Model: The specific model file to use.
  • Speaker ID (VITS only): The voice to use for the narrator.
  • Models Folder: The folder where local speech models are stored.
  • Speech Localization: Choose whether to synthesize speech from the translated text or the original English text.
  • Keep Model In Memory: Keeps the speech model loaded in RAM.

Speech Recognition

These settings control the speech-to-text for voice input.

  • Provider: Choose between OpenAI (cloud), Local, or LocalFast.
  • Microphone: Select your active microphone.
  • Auto Submit: Automatically send your recorded speech.

Local Speech Recognition Settings

  • Models Folder: The folder where local speech recognition models (Whisper) are stored.
  • Model: The specific .bin model file to use.
  • Keep Model In Memory: Keeps the recognition model loaded in RAM/VRAM.
  • Use GPU: Offload processing to the GPU for faster recognition.

Cloud Speech Recognition Settings

  • Model: Select the speech recognition model offered by your chosen cloud provider.