fal-ai

Run generate_image.py with uv, passing a prompt, output path, and model flag such as -m recraft-v4. The script requires the FAL_KEY environment variable and supports aspect ratio, custom dimensions, seed, and multiple images.

Recraft-v4 is the recommended model for logos, banners, and design assets at roughly $0.04 per image. For scalable SVG vector output, use recraft-vector, and for the highest design quality use recraft-v4-pro.

Flux2-klein is the cheapest option at about $0.01 per image with sub-second generation. Fast-lightning-sdxl is another ultra-fast budget option for rapid iteration on game assets and prototypes.

Run generate_image.py with the --usage flag, optionally adding --usage-days to set the date range. This requires the FAL_KEY_ADMIN environment variable with an admin key from the fal.ai dashboard, since the usage API rejects regular keys.

Yes, pass any endpoint directly with the --endpoint flag, such as fal-ai/custom/model. The Skill also offers a Serper search fallback to look up unfamiliar models on fal.ai before generating.

The Skill enforces a mandatory confirmation step showing the exact prompt, model, settings, and estimated cost. Generation only proceeds after you explicitly type Y or yes, preventing accidental API charges.