ai-image-gen（AI 图像生成）

AI Image Generation Skill

Generate or edit images using the ai-img CLI with Vercel AI SDK.

When to use

• Generate images from text prompts
• Edit existing images with AI (inpainting, style transfer)
• Batch process multiple image generation jobs

Prerequisites

1. Install CLI: # Option 1: Install globally with Bun (recommended) bun add -g ai-img # Option 2: Clone and build git clone https://github.com/aryasaatvik/ai-img cd ai-img bun install bun run build
2. Set API key (choose one):

- OpenAI: export OPENAI_API_KEY=sk-... - Google: export GOOGLE_API_KEY=..., export GEMINI_API_KEY=..., or export GOOGLE_GENERATIVE_AI_API_KEY=... - FAL AI: export FAL_API_KEY=...

Quick Decision Tree

Image Generation?
├─ Generate new image → ai-img generate
├─ Edit existing image → ai-img edit
└─ Many images at once → ai-img batch

Commands

generate

Generate new images from text prompts.

# Basic
ai-img generate --prompt "A red cat"

# Multiple images
ai-img generate --prompt "A sunset" --count 3

# Specific provider/model
ai-img generate --prompt "A cat" --provider fal --model "fal-ai/flux/dev"

# With size and quality
ai-img generate --prompt "Landscape" --size 1536x1024 --quality high

edit

Edit existing images with AI.

# Basic edit
ai-img edit --input photo.jpg --prompt "Make it vintage"

# With mask (partial edit)
ai-img edit --input photo.jpg --mask mask.png --prompt "Add sunglasses"

# Multiple input images
ai-img edit --input "product.jpg,background.jpg" --prompt "Place product in scene"

batch

Process multiple jobs from JSONL file.

# Create jobs file (one JSON per line)
echo '{"prompt": "A red car", "out": "red-car.png"}
{"prompt": "A blue car", "out": "blue-car.png"}' > jobs.jsonl

# Run batch
ai-img batch --input jobs.jsonl --outDir ./output --concurrency 5

Options Reference

Providers & Models

See references/providers.md for full model list.

Prompting Tips

• Be specific about style: "photorealistic", "oil painting", "3D render"
• Include composition: "centered", "wide shot", "close-up"
• Specify mood/lighting: "golden hour", "dramatic shadows", "soft light"
• Add constraints: "no text", "no logos"

See references/prompting.md for detailed guidance.

Examples

Product Photography

ai-img generate --prompt "White ceramic mug on wooden table, product photography, soft natural light, clean background" --provider openai --size 1024x1024 --quality high

Icon Generation

ai-img generate --prompt "Simple geometric rocket icon, flat design, blue and white, minimal" --provider fal --model "fal-ai/flux/dev" --size 512x512

Background Replacement

ai-img edit --input product.jpg --prompt "Replace background with warm sunset gradient, keep product unchanged" --provider openai

Troubleshooting

• "Missing API key": Ensure your API key is set in the environment
• "Invalid model": Check the model ID is valid for the provider
• Rate limiting: Reduce --concurrency for batch jobs

See references/cli.md for full CLI reference.