Image Character Consistency and Composition Skill
This skill outlines a robust workflow for performing various image manipulations while ensuring the central character remains consistent in appearance, style, and proportions. It also covers compositional adjustments to facilitate multi-character scenes or specific framing needs.
Core Principles for Image Modification
When modifying images involving characters, it is crucial to maintain consistency across all generated outputs.
This includes:
Character Appearance: Facial features, hair color and style, skin tone, and overall artistic rendering should remain faithful to the original character.
Proportions and Perspective: Any changes to pose or body parts must respect the character's inherent proportions and the perspective of the scene to avoid distortion.
Art Style and Palette: The generated image should seamlessly blend with the original art style and color scheme, especially when integrating elements from different source images.
Safety Compliance: Be mindful of content policies. When describing clothing or poses, use neutral and professional language to avoid triggering automated safety filters.
If a request is flagged, rephrase it to be less suggestive while still conveying the desired aesthetic.
Workflows for Image Manipulation
This skill provides guidance for the following common image manipulation tasks:
1. Head Replacement/Swapping
To replace a character's head from one image with another, use the generate_image_variation tool.
The prompt should explicitly state the source and target images, emphasizing the need to maintain proportions, angle, perspective, and color palette.
Example Prompt:
Plain Text
A professional digital illustration that seamlessly replaces the head of the character in `[Image 1 path]` with the head from `[Image 2 path]`, ensuring perfect alignment of proportions, angle, perspective, and color palette.
Maintain the original art style and lighting.
2. Outfit Changes
To modify a character's clothing, describe the new attire in detail within the generate_image_variation prompt.
Be precise about each clothing item, its color, style, and fit. If previous attempts were flagged by safety filters, use neutral descriptions.
Example Prompt (Neutral):

name: image-character-consistency
description: Provides workflows for maintaining character consistency across multiple image generations, including head swaps, outfit changes,
pose adjustments, and compositional shifts. Use this skill when modifying existing character images or generating new ones based on a consistent character.
license: Complete terms in LICENSE.txt

A full-body digital illustration in a watercolor-like art style. The character is wearing a classic white button-up blouse, an elegant black knee-length skirt,
and black high-heeled shoes. Maintain the character's appearance, pose, and the background from `[Reference Image Path]`.

Tool Usage
generate_image_variation: Ideal for modifying existing images, such as changing outfits or performing head swaps, as it uses a reference image to maintain consistency.
generate_image: Use when generating a new image from scratch, such as a new pose in a new environment, while still referencing previous images for character consistency.
Always use the references parameter in generate_image or generate_image_variation to link back to previous successful generations, ensuring visual continuity. Prioritize clear, detailed, and neutral language in prompts to achieve desired results and comply with safety guidelines.