SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsimage-gen

image-gen

Design & creation

View repository

Updated 10-15 22:09

About this skill

Image Gen — Integrated AI Image Generation

Overview

image-genis the integrated image generation skillof the moai-media plugin. It analyzes the user's requirements and automatically routes to the optimal AI model.

  • Higgsfield Soul (default): cinematic images, product photography, editorial style. Supports 15+ style presets.
  • fal.ai models: Flux 1.1 Pro (photorealism), Ideogram 3.0 (Korean typography), Recraft V3 (vector/logo).
  • Gemini Direct: special case for Korean typography.

Trigger keywords

This skill is automatically invoked for requests such as the following:

  • "Generate an image", "Make an image"
  • "Card news image", "Instagram thumbnail"
  • "Product photo", "detail page banner"
  • "Poster design", "ad creative"
  • "Make a logo", "vector icon"
  • "Cinematic image", "movie-like photo"
  • "/moai-media image-gen" (direct invocation)

Workflow

Step 1: Analyze user requirements

Identify the following from the user prompt:

  • Whether text is included: whether there is Korean/English typography
  • Style: cinematic, product photography, editorial, photorealism, vector/logo
  • Aspect ratio: 1:1, 3:4, 9:16, 16:9 등

Step 2: Model routing

Situation Optimal model Routing criteria
Large Korean typography Ideogram 3.0 (fal.ai) or Gemini Direct Text occupies 30% or more of the image
Cinematic/product photography Higgsfield Soul (default) product_photography, cinematic, editorial styles
General photorealism Flux 1.1 Pro (fal.ai) General photographic realism
Vector/logo Recraft V3 (fal.ai) Logo, icon, vector graphics
Small Korean typography Gemini Direct When text is less than 10%
Default Higgsfield Soul No explicit style specified

Step 3: Image generation

Call the selected model's MCP tool or Python script.

Usage examples

Example 1: Cinematic product photo (Higgsfield Soul)

"프리미엄 커피 머신 제품 사진, 편집지 스타일, 부드러운 조명, 미니멀 배경"
→ Higgsfield Soul, style_preset: product_photography

Example 2: Card news title (Ideogram 3.0)

"'2025년 봄 트렌드'라는 한글 제목, 파스텔 톤 배경, 인스타그램 세로"
→ Ideogram 3.0 via fal.ai, aspect_ratio: 3:4

Example 3: Brand logo (Recraft V3)

"미니멀 로고, 서울 브랜드, 깔끔한 벡터 스타일, 블루와 화이트"
→ Recraft V3 via fal.ai

Example 4: Photorealistic landscape (Flux Pro)

"서울 여의도 공원 봄 벚꽃, 해 질 녘, 따뜻한 조명"
→ Flux 1.1 Pro via fal.ai

Output format

Higgsfield Soul (MCP)

MCP 툴: higgsfield.generate_image_soul

Main parameters:

  • prompt (string): image description
  • style_preset (string): style preset (product_photography, editorial, cinematic, abstract, minimalist, etc. 15+)
  • aspect_ratio (string): 1:1, 3:4, 9:16, 16:9
  • num_images (int): number of images to generate (1-4)
  • seed (int, optional): Seed value for reproducibility

Output: Array of image URLs

fal.ai Models (MCP)

MCP 툴: fal-ai.run_inference

Parameters by model:

  • Ideogram 3.0: fal-ai/ideogram/v3 - prompt, rendering_speed (TURBO/DEFAULT/QUALITY), aspect_ratio
  • Flux 1.1 Pro: fal-ai/flux-pro/v1.1-ultra - prompt, image_size, num_inference_steps
  • Recraft V3: fal-ai/recraft-v3 - prompt, style_vector, aspect_ratio

Gemini Direct (Python script)

python "${CLAUDE_PLUGIN_ROOT}/scripts/generate_image.py" \
  "프롬프트" output/image.png 3:4 nano-banana-pro

Notes

API key required

To use this skill, the following API keys are required:

Service Environment Variable Issuer Cost
Higgsfield HIGGSFIELD_API_KEY, HIGGSFIELD_SECRET Higgsfield AI ~$0.05-0.15/장
fal.ai FAL_KEY fal.ai/dashboard/keys Varies by model
Gemini GEMINI_API_KEY ai.google.dev $0.067-0.134/장

Cost Reference

Model Purpose Cost per image
Higgsfield Soul Cinematic/Product Photography ~$0.05-0.15
Ideogram 3.0 Korean Typography ~$0.03-0.08
Flux 1.1 Pro Photorealism ~$0.04
Recraft V3 Vector/Logo ~$0.02-0.05
Gemini Pro Korean Typography Special $0.134 (2K)
Gemini Flash Korean Typography Special $0.067 (1K)

Usage Recommendations

  • Bulk Generation: Use Gemini Flash or Ideogram TURBO mode for cost optimization
  • Final Delivery: Use Higgsfield Soul QUALITY or Ideogram QUALITY mode
  • A/B Testing: Create drafts with cheaper models, then generate final versions with premium models

Related Skills

  • nano-banana (moai-media) — Gemini 3 Image skill specialized in Korean typography
  • video-gen (moai-media) — Integrated video generation (image→video, text→video, fal.ai Kling·Hailuo backend)
  • audio-gen (moai-media) — TTS·voice synthesis (uses ElevenLabs MCP)
  • speech-video (moai-media) — Voice+video lip-sync combination
  • character-mgmt (moai-media) — Higgsfield character management (brand consistency)
  • fal-gateway (moai-media) — Integrated entry point for 1000+ fal.ai models including Flux·Ideogram·Recraft

References