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

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

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

| Flag | Description | Default | | ---------------- | -------------------------------------- | ---------------- | | --prompt, -p | Text prompt | Required | | --provider, -P | AI provider: openai, google, fal | openai | | --model, -m | Model ID | Provider default | | --size, -s | Image size (WxH) | 1024x1024 | | --aspectRatio | Aspect ratio (e.g., 16:9) - for Gemini | - | | --count, -c | Number of images | 1 | | --seed | Random seed | Random | | --quality | Quality: low/medium/high/auto | auto | | --output, -o | Output file path | output.png | | --outDir | Output directory | Current dir |

Providers & Models

| Provider | Default Model | Environment Variable | | -------- | ----------------------- | -------------------------------- | | OpenAI | gpt-image-1.5 | OPENAI_API_KEY | | Google | imagen-3.0-generate-002 | GOOGLE_API_KEY or GEMINI_API_KEY | | FAL AI | fal-ai/flux/dev | FAL_API_KEY |

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.