Parameter Glossary
This glossary defines common parameters you'll encounter when configuring AI models in Fater. These settings allow you to fine-tune the generation process and are typically found in the Left Sidebar after selecting an AI model.
For detailed ranges, defaults, and specific behaviors for each AI model, please refer to the AI Model Directory.
Common Parameters:
Prompt:
Definition: The primary text description you provide to the AI, detailing the subject, scene, style, actions, and other attributes you want in the generated output.
Applies to: Most AI models (Image, Asset, Video generation & editing).
(See also: Text Prompts & Negative Prompts)
Negative Prompt:
Definition: A text description of elements, styles, or qualities you want the AI to avoid generating.
Applies to: Many AI models.
(See also: Text Prompts & Negative Prompts)
Seed:
Definition: A numerical value that initializes the random generation process. Using the same seed with identical prompts and parameters will produce the same result. A value of
-1
typically means a random seed.Applies to: Almost all AI generation models.
(See also: Seeds)
Steps (Diffusion Steps / Iterations):
Definition: Generally controls the number of refinement iterations the AI performs. More steps can lead to higher detail and coherence but increase generation time.
Applies to: Many image generation, editing, and upscaling models. The optimal range varies significantly per model.
Guidance Scale (CFG Scale / Prompt Adherence):
Definition: Determines how strictly the AI should follow your text prompt. Higher values mean closer adherence; lower values allow more creative deviation.
Applies to: Many image and video generation/editing models. The effective range and impact vary per model.
Number of Generations:
Definition: Specifies how many output variations the AI should produce from a single click of the "Generate" button, using the same core parameters (but typically different random seeds if the main Seed is set to -1).
Applies to: Primarily Image Editor models, some Asset/Video models.
Parameters Common in Image Editing & Inpainting Models:
Style Type:
Definition: Allows selection from predefined artistic styles (e.g., Realistic, Anime, Design) to influence the generated content. "AUTO" often lets the AI decide.
Destroy Fill Area Behind Mask:
Definition: A boolean (Yes/No) setting. "Yes" means the AI completely ignores original pixels under a mask; "No" allows them to influence the inpainting. Relates to Denoising Strength.
Denoising Strength (or Image Strength / Creativity for Inpainting):
Definition: Controls how much the AI alters existing pixels within a masked area (or a source image for image-to-image tasks). A value of 1.0 usually means full replacement; lower values preserve more of the original.
Mask Alpha Radius (Edge Blending):
Definition: Determines the smoothness of the transition (transparency feathering) between AI-generated content and the original image at the mask's edges.
Mask Extra Radius (Generation Padding):
Definition: The AI generates content slightly beyond the defined mask by this many pixels, which can help with seamless blending.
Parameters Common in Upscaling Models:
Scale (Factor):
Definition: A direct multiplier for the image dimensions (e.g., 2x, 4x).
Output Megapixels:
Definition: The target final resolution of the upscaled image, specified in millions of pixels.
Creativity (for Upscaling):
Definition: Controls how much new detail the AI is allowed to invent or hallucinate during the upscaling process.
Resemblance:
Definition: Determines how faithfully the upscaled image should stick to the forms and structures of the original low-resolution image.
Sharpen:
Definition: Applies an additional sharpening effect.
Dynamic (HDR):
Definition: Influences the High Dynamic Range of the output.
Tile Size:
Definition: An internal processing parameter affecting how the image is divided for upscaling. Can impact performance and potential artifacts.
Hand Fix:
Definition: A boolean (Yes/No) to activate a specialized module for improving the rendering of hands and fingers.
Checkpoint:
Definition: Allows selection of a specific underlying base model or training version for the upscaler.
Parameters Common in Video Generation Models:
Duration:
Definition: Sets the length of the generated video clip in seconds.
This glossary provides general definitions. For the exact behavior, recommended ranges, and default values for parameters of a specific AI model, please consult its entry in the AI Model Directory.
Last updated