jdy-imagine

name	jdy-imagine
description	AI image generation via Google Gemini, OpenAI gpt-image-2, and apimart (China gateway). Text-to-image, image-to-image, edit with mask, batch generation at 50% cost (Google/OpenAI), 4K + 13-value ar (apimart).

jdy-imagine

AI image generation plugin for Claude Code. Supports Google Gemini, OpenAI gpt-image-2, and apimart (China-domestic gateway for gpt-image-2 with native 4K + 13-value ar).

Usage

Text-to-image

bun scripts/main.ts generate --prompt "A cat in watercolor style" --outdir ./images
bun scripts/main.ts generate --provider openai --prompt "A cat in watercolor style" --outdir ./images

Image-to-image (reference)

bun scripts/main.ts generate --prompt "Make it blue" --ref source.png --outdir ./images

Edit (with mask, OpenAI / apimart)

bun scripts/main.ts generate --provider openai \
  --prompt "Replace background" --edit photo.png --mask mask.png --outdir ./images

apimart (China gateway, 4K, extended ar)

# Native 4K (only ar 16:9 / 9:16 / 2:1 / 1:2 / 21:9 / 9:21)
bun scripts/main.ts generate --provider apimart \
  --prompt "..." --resolution 4k --ar 16:9 --outdir ./images

# Apimart-only aspect ratios (5:4, 4:5, 2:1, 1:2, 21:9, 9:21)
bun scripts/main.ts generate --provider apimart --prompt "..." --ar 21:9 --outdir ./images

Batch generation (50% cost savings)

bun scripts/main.ts batch submit prompts.json --outdir ./images
bun scripts/main.ts batch submit text-only-prompts.json --provider openai --outdir ./images

(Note: command must come before flags. --provider openai batch submit ... will not parse — --provider would be consumed as a flag and command would be empty.)

Options

• --provider: google (default), openai, or apimart
• --model, -m: Model ID (provider default if not specified)
• --ar: Aspect ratio. CLI accepts 13 values: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 (google + openai), plus 5:4, 4:5, 2:1, 1:2, 21:9, 9:21 (apimart-only when added — google + openai reject these via validateRequest)
• --resolution: 1k / 2k / 4k (default: 2k; 4k apimart-only when added — google + openai reject)
• --detail: auto / low / medium / high (default: high; OpenAI passes through, Gemini ignores)
• --ref: Reference image path(s) — works in both providers
• --edit: Edit target image path — Google: same as --ref; OpenAI: routes to /edits
• --mask: Mask image path — OpenAI / apimart, requires --edit or --ref (Google rejects)
• --outdir, -o: Output directory (default: .)
• --json: JSON output mode

Configuration

• Google: GOOGLE_API_KEY or GEMINI_API_KEY
• OpenAI: OPENAI_API_KEY
• apimart: APIMART_API_KEY

Or create .jdy-imagine/.env.

Capability matrix

Feature	Google	OpenAI	apimart
--ref	yes (inlineData)	yes (routes to /edits)	yes (auto-uploads to /v1/uploads/images, 72h URL)
--edit	falls back to --ref	yes (native)	yes (image_urls[0])
--mask	not supported	yes (needs --edit or --ref)	yes (mask_url field)
--chain	yes	not supported	not supported
--character	yes	yes (realtime); blocked in OpenAI batch	yes (sha256 cache dedupes uploads)
batch submit	yes	text-only (no refs/edit/mask/character)	not supported (submit/poll already async)
4K	no	not exposed	yes — only with ar 16:9 / 9:16 / 2:1 / 1:2 / 21:9 / 9:21
Extended ar (5:4, 4:5, 2:1, 1:2, 21:9, 9:21)	rejected	rejected	yes
HTTPS_PROXY	yes	text-only	text-only (refs require disabling proxy)