| name | generate-images |
| description | Generate one or more images from a text prompt using OpenAI's Images API (model gpt-image-1.5). Use when the user asks to “generate images”, “create an image”, “make a thumbnail”, or similar; especially when they provide a prompt plus options like count/size/quality/output directory. |
Generate Images (gpt-image-1.5)
Generate images via the OpenAI Images API using a deterministic helper script.
Prerequisites
OPENAI_API_KEY environment variable must be setjq must be available (used to parse API JSON)
Usage
/generate-images a sunset over the ocean — generate 1 image/generate-images 3 cats wearing hats — generate 3 images (shorthand count)/generate-images --count 5 --size 1536x1024 a mountain landscape — full options
Arguments
The argument string is available as $ARGUMENTS.
Flags (all optional)
--count N (default 1) — number of images (1–10)--size WxH (default 1024x1024) — one of: 1024x1024, 1536x1024, 1024x1536--quality Q (default high) — low, medium, high--output DIR (default ./) — directory to save images--prefix NAME (default image) — filenames like {prefix}-1.png, {prefix}-2.png, ...--format FMT (default png) — png or jpg
Flags can appear in any order and may be mixed with prompt tokens.
Shorthand count
If the prompt starts with a bare number + space (e.g. 3 cats wearing hats), treat the number as the count only if no --count flag is present.
Everything else is the prompt
After extracting flags and optional leading count, join the remaining text as the prompt.
Procedure
Prefer running the bundled script (it handles parsing, file naming, progress, and errors):
bash /Users/ericblue/clawd/skills/public/generate-images/scripts/generate-images.sh
The script reads $ARGUMENTS (or you can pass a single argument string as $1).
Notes
- Each API call generates one image; multiple images are generated sequentially.
- Cost varies by quality and size (rough estimates per 1024×1024 image): low ~$0.04, medium ~$0.07, high ~$0.19. Larger sizes are ~1.5×.
