Glossary
Definitions of terms used in VariantLab.
A
Avatar
Your profile picture, shown in the app header, sharing dialog, and member lists. Upload from Account Settings or revert to initials.
Aspect Ratio
The proportional relationship between an image's width and height (e.g., 1:1 for square, 16:9 for widescreen).
B
Base Image
The original AI-generated or uploaded image that serves as the foundation for your project. You can have up to 5 base images per project.
Bounding Box
The rectangular coordinates defining the position and size of a mask or layer within the full image.
Brush Tool
The drawing tool used to add areas to a mask (paint white pixels).
C
Combination
A unique composite image created by selecting one layer from each trait and combining them.
Crop to Content
An export option that trims transparent edges from images, removing empty space.
D
Detection
The AI process of identifying and creating masks for specified traits in an image.
Detection Prompt
Text that tells the AI what to look for when creating a mask (e.g., "the robot's large round eyes").
Display Name
A human-readable identifier for images used in exports (e.g., final_eyes_v0_hat_v2.png).
E
Eraser Tool
The drawing tool used to remove areas from a mask (paint black pixels).
Export
The process of downloading your images as a ZIP file in your chosen format.
F
Favorite
A starred item (base image, mask variant, or layer) that is selected for use in the pipeline.
Flash Model
Gemini 2.5 Flash Image - the fast, cost-effective AI model for image generation.
Fork
Creating your own copy of a template, including its configuration and optionally its images. See Template Gallery.
G
Gemini
Google's family of AI models used by VariantLab for image generation and analysis.
L
Layer
A transparent PNG image extracted from a variation, ready to be combined with other layers.
Layer Extraction
The automatic process of creating a transparent layer from a variation using its trait's mask.
Layer Order
The back-to-front arrangement of traits determining how layers stack in combinations.
M
Mana
VariantLab's credit currency used for AI operations like image generation, trait detection, and export.
Mask
A black-and-white image defining which pixels belong to a trait. White = included, black = excluded.
Mask Variant
One of up to 5 versions of a mask for a single trait. Allows trying different detections or manual alternatives.
O
Original Layer
The layer extracted from the base image for a trait (before any variations).
P
Pipeline
The 5-stage workflow: Setup → Detection → Variations → Combine → Export.
Pro Model
Gemini 3 Pro Image Preview - the high-quality AI model supporting up to 4K resolution.
Project
A container for all your work: base images, masks, variations, and combinations.
Prompt
Text describing what you want the AI to generate or detect.
Q
Quality Level
Output quality setting (low, medium, high) available on some models. Higher quality produces more detailed images but costs more Mana. See AI Image Generation.
R
Role
Your permission level in a shared project: Owner, Editor, or Viewer. See Project Sharing.
S
Slot
A position for a base image (1-5) in the Setup page.
T
Template
A published project that other users can discover in the gallery and fork to create their own version. See Template Gallery.
Trait
A distinct part of your image that can be varied independently (e.g., eyes, hat, background).
Thumbnail Stack
The vertical strip of thumbnails for navigating between variants.
V
Variant
One version of a mask or layer. Each trait can have up to 5 mask variants and 5 variants per variation prompt.
Variation
An AI-generated modification of a specific trait based on a prompt.
Variation Prompt
Text describing how to modify a trait (e.g., "glowing red LED eyes").
Z
ZIP Export
The downloadable package containing all your exported images and optional metadata.