Overview
A character management skill for brand consistency in AI content creation. Using the Higgsfield AI character system, create brand mascots, virtual influencers, product endorsers, etc. from just 2–5 reference images and a description, and reuse them across all Higgsfield tools such as image generation (Soul), video generation (DOP), and speech video character_idis issued.
Why is character management needed?
- Brand consistency: Maintain the same character across marketing campaigns, social media, and advertising
- Time saving: No need to repeat the character description in the prompt every time
- Quality assurance: Output of similar quality every time through learning based on reference images
- Scalability: Can generate 100+ pieces of content with a single character
Trigger keywords
Use this skill for the following requests:
- Related to character creation: "Create a character", "brand mascot", "virtual influencer", "character design", "AI avatar"
- Character reuse: "Find the character I made before", "character_id lookup", "character archiving", "saved character"
- Character modification/deletion: "Delete character", "regenerate mascot", "character list"
- Branding: "brand face", "product mascot", "series character"
Workflow
1. Character creation
1. 참조 이미지 준비 (2~5장, 정면/측면/다양한 포즈)
2. 캐릭터 설명 작성
- 이름: 예) "Luna the Tech Cat"
- 외형: 예) "회색 고양이, 넥타이, 안경"
- 성격: 예) "친근, 전문적, 유머러스"
- 스타일: 예) "3D 렌더링, 픽사 스타일"
3. create_character 실행 → character_id 발급
4. 다른 스킬에서 character_id로 재사용
2. Character search
1. get_character에 character_id 전달
2. 캐릭터 상세 정보 반환 (이름, 설명, 생성일)
3. Character deletion
1. delete_character에 character_id 전달
2. 영구 삭제 (복구 불가)
Usage examples
Example 1: Create a brand mascot
"우리 스타트업 'TechNova'의 브랜드 마스코트를 만들어줘.
이름: Nova the Fox
외형: 주황색 여우, 하얀 가운, 고글 착용
성격: 호기심 많은 연구원
스타일: 3D 애니메이션, 귀여운 분위기
참조: fox_ref_1.png, fox_ref_2.png, fox_ref_3.png"
→ 출력: character_id = "characters/abc123"
Example 2: Create a virtual influencer
"뷰티 유튜브 채널용 가상 인플루언서를 만들어줘.
이름: Sofia
외형: 20대 후반 한국인 여성, 긴 웨이브 헤어
스타일: K-뷰티, 세련된 메이크업
성격: 친근, 전문적, 트렌디
참조: sofia_front.jpg, sofia_side.jpg, sofia_smile.jpg"
→ 출력: character_id = "characters/def456"
Example 3: Generate an image with a created character
(이전에 생성한 Nova the Fox 캐릭터 사용)
"Nova the Fox 캐릭터로 AI 도구 소개 이미지 5장 만들어줘.
1. 코드를 작성하는 Nova (노트북 앞)
2. 발표하는 Nova (프레젠테이션)
3. 커피를 마시는 Nova (휴식)
4. 팀원들과 협업하는 Nova (화상 회의)
5. 축하하는 Nova (성공)"
→ image-gen 스킬에서 character_id = "characters/abc123" 사용
Example 4: Search for and delete a character
"저장한 캐릭터 목록 보여줘.
character_id가 'characters/old_mascot'인 캐릭터는 삭제해줘."
→ get_character로 조회 후 delete_character로 삭제
Output format
Character creation output
{
"character_id": "characters/abc123",
"name": "Nova the Fox",
"description": "주황색 여우 마스코트, TechNova 브랜드",
"created_at": "2026-04-30T10:00:00Z",
"status": "active"
}
Character search output
{
"character_id": "characters/abc123",
"name": "Nova the Fox",
"description": "주황색 여우 마스코트, TechNova 브랜드",
"reference_images": ["image_1.png", "image_2.png"],
"created_at": "2026-04-30T10:00:00Z"
}
Character deletion output
{
"character_id": "characters/abc123",
"status": "deleted",
"deleted_at": "2026-04-30T12:00:00Z"
}
Notes
API key required
HIGGSFIELD_API_KEY + HIGGSFIELD_SECRET environment variable is required.
- higgsfield.com Sign up (currently in beta)
- Settings → API Keys → Create Key Pair
.envor register it in a system environment variable:export HIGGSFIELD_API_KEY="your_api_key" export HIGGSFIELD_SECRET="your_secret_key"
Reference image guide
Requirements for a good reference image:
- At least 2, at most 5: Various angles and expressions
- Front-facing image required: Clearly identify facial features
- High resolution: At least 512x512 pixels
- Consistent style: All reference images have similar style/lighting
- Simple background: So the character is clearly distinguished
Examples of bad reference images:
- Blurry or noisy images
- Extremely different styles (e.g., 1 photo + 1 cartoon)
- When there are only images with faces obscured
- Overly filtered images
Character consistency scope
Generated characters are only within the Higgsfield tool consistency is guaranteed:
- ✅ Supported: image-gen (Soul), video-gen (DOP), speech-video
- ❌ Not supported: fal-gateway, skills of other plugins
To use characters outside of Higgsfield:
- First generate an image containing the character in Higgsfield
- Use that image as a reference image for other tools (e.g., IP Adapter)
Pricing and limits
- Beta period: Free (policy subject to change)
- Number of saved characters: Unlimited (during beta period)
- Deletion caution: Deleted characters cannot be recovered
Character ID management tips
- Naming convention:
characters/{brand}_{name}format recommended- 예:
characters/technova_nova,characters/sofia_beauty
- 예:
- Documentation: Record character_id and purpose in project documentation
- Version management: Create a new ID when modifying a character (overwriting not possible)
Related skills
Character consumption skills
- image-gen: Generate images containing the character (Soul model)
- video-gen: Generate videos featuring the character (DOP model)
- speech-video: Generate videos of the character speaking (lip sync)
Complementary skills
- moai-content:copywriting: Define character personality, dialogue, and storyline
- moai-domain-brand-design: Define character style that fits brand guidelines
MCP server settings
This skill uses Higgsfield MCP (stdio).
{
"higgsfield": {
"command": "uvx",
"args": ["higgsfield-mcp"],
"env": {
"HIGGSFIELD_API_KEY": "${HIGGSFIELD_API_KEY}",
"HIGGSFIELD_SECRET": "${HIGGSFIELD_SECRET}"
}
}
}
MCP server registration procedure: Refer to moai-media/CONNECTORS.md .
Character lifecycle
graph LR
A[참조 이미지 수집] --> B[캐릭터 생성]
B --> C[character_id 발급]
C --> D[이미지 생성]
C --> E[비디오 생성]
C --> F[스피치 비디오]
D --> G[콘텐츠 배포]
E --> G
F --> G
G --> H[성과 측정]
H --> I[캐릭터 수정 필요?]
I -->|예| J[새 참조 이미지]
J --> B
I -->|아니오| C
Key: A character created once branches out into numerous content. The initial reference image investment is important.