How to Create Filter

The Create Filter page is where you create, test, refine, and finalize filters before publishing them.

This page provides an overview about:

1. Choose the correct model

Inside the Filter Creator you can work with all our available models.

Model	Supported Persons	Available Transformation logics	Limitations
FaceSwap v5	1 Person	Text Filter Inpainting	IP-related assets can be enabled via inpainting mode
FaceSwap v6	1 Person Soon: Multi-Person (for up to 4 persons)	Text, Inpainting (Img2Filter), Soon: Generative (Text + Assets)	IP-related assets can be enabled via inpainting mode
MultiSwap (Stylized & Realistic)	1–5 Persons	Text only	IP-related assets cannot be used (e.g. integration of a character, a object or similar)

Understanding the Transformation Logics

To fully leverage the Filter Creator, it is essential to understand the three filter logic types.

Category	🟢 Text Filter	🟢 Inpainting Filter	🟢 Generative Creation (Text + Assets)
Definition	Filter is created purely via a structured text prompt.	A predefined concept is activated via masking + text logic.	Fully generative filter combining structured assets with text-driven scene composition.
How It Works	Prompt defines the full scene transformation No predefined assets required AI generates output entirely from prompt logic	Define areas to modify via masking Enable fully body or only partial transformations (only visible skin areas) Background & non-masked areas stay controlled	Combine multiple created assets (jersey, background, objects) • Use text prompt to describe final scene • AI generates a fully new image • Defined assets remain visually consistent
Level of Control	Low	Very High	Very High
What Stays Stable	no elements are consistent	everything not masked	Uploaded Assets,
What Changes	Clothing, environment, style, lighting	Masked areas only (e.g. outfit, body)	Full scene composition with preserved assets
Ensures	Fast execution & creative flexibility	Partial, precise transformation Stable composition High photorealistic consistency	Maximum creative flexibility Asset consistency without detail loss Complex scene composition
Best For	Fun Purposes, enables to create unspecific scenes in a fully generative way	Any kind of serious activation requiring consistent details for specific assets Branded outfit swaps Jersey replacements Controlled body transformations Maintaining background consistency	Brand-heavy campaigns Multi-asset activations Premium photorealistic transformations
Model Availability	V5, V6, MultiSwap	V5, V6	V6 (extended asset support coming soon)

2a. Creating & Testing via Text only (Text 2 Filter)

Creating with our Faceswap Models are possible in two ways:

A. Text 2 Filter (via Prompting) - (supports both Faceswap or Multiswap)

Write your prompt describing the final output.

Wire a prompt or use the Prompt Suggestion to auto-generate and refine your prompt.
Upload a test person to preview results
Adjust settings (aspect ratio, prompt settings, etc.)
Generate and iterate until satisfied