Prompting & Masking Guide Img2Filter: Faceswap V5

In this chapter you learn, how to:

How to Prompt Img2Filter (Faceswap V5)

V5 uses descriptive prompts that match the base image. Focus on visual details and style, while keeping the scene aligned with the reference image.

Example prompt (realistic):

photo of confident person in action movie style wearing jungle outfit, angry expression, looking at the camera, cinematic lighting, high quality

Example negative prompt:

helmet, hat, sunglasses, weapon

Do’s for Prompting:

Keep prompts clear and concise
Match the prompt to the base image
Use negative prompts to remove unwanted elements
Avoid changing the camera perspective or pose
Use keywords (lighting, style, mood)

Don't Promp:

Always align your prompt with what is already visible in the base image.

Indirect phrasing
❌ astronaut without helmet
✅ astronaut + negative: helmet
Too many ideas in one prompt
Changing camera angle drastically from base image

How to Mask:

Masks should be rough, general shapes
Avoid tracing the subject exactly
Keep it loose and soft-edged so filters can adapt to different body types and poses
The goal is to define a general region, not a precise cutout

Advanced Settings for Faceswap V5:

Denoise strength (0–1.0)

Controls how much the output deviates from the input image.

· Values 1-0.7 → average changes (recommended)

· Lower values preserve more original detail but likness of user can be too low

Depth strength (0–1.0)

Controls how strictly shape/pose is maintained.

· Very high values (0.7–1.0) reduce flexibility

· Recommended: 0.2–0.4

Hand deformity fixes

If results show hand issues:

· Slightly lower denoise (but not below 0.5)

· Increase depth strength

· Use input images with gloved or partially obscured hands

Masking Options & Example Prompts for V5:

Category	Example 1: Fully Body Mask	Example 2: only mask Skin & Head	Example 3: only mask Head
Masking Reasoning	Full body mask since no necessary details on the body need to stay identical (e.g., no jersey or fixed outfit elements).	The body fits well in the scenery and appears unisex, therefore no full body mask required (=head + visible skin areas)	The armor must stay identical, therefore only visible skin areas are masked. If this filter should support females/children, additional base images are required.
Prompt V5	photo of confident person in action movie style, wearing a dirty grey tank top and jungle attire, angry expression, looking at the camera, high quality	photo of a confident person, confident expression, looking at the camera, high quality	photo of a confident person wearing a astronaut helmet, focused expression, looking at the camera, high quality, 16k,
Benefit	All Users have different clothes and slightly different poses.	The armor stays exactly the same	Due to the perspective no other
Input image
Example for Mask
Result V5