| name | openrouter-imagegen |
| description | Generate images cost-effectively via OpenRouter (Seedream 4.5, FLUX.2, Recraft V4, etc.) as a cheaper alternative to OpenAI gpt-image. Use when the user wants low-cost image generation, a specific OpenRouter image model, or quality comparable to gpt-image-2 at a fraction of the price. Requires OPENROUTER_API_KEY in .env.local. |
OpenRouter Image Generation (cost-effective)
A unified gateway to many image models. Use for the cheap-but-high-quality path,
or to reach models only available via OpenRouter (Seedream, FLUX, Recraft, Sourceful).
Prerequisites
OPENROUTER_API_KEY in .env.local (placeholder present — paste your key from
https://openrouter.ai/keys). Until set, the helper exits with a clear message.- Helper script:
scripts/generate-image-openrouter.ts (no extra npm deps — global fetch).
⚠️ Two facts that matter
- No Qwen image model on OpenRouter (checked live). Qwen-Image is served via Alibaba
DashScope, not OpenRouter. Don't promise it here.
- Gemini image models are cheaper called DIRECTLY via
GEMINI_API_KEY (no OpenRouter
markup) — use the nb-* skills for those. Use this skill for Seedream / FLUX / Recraft /
OpenAI gpt-5-image, which you can only reach via OpenRouter.
Cost-effective model picks (≈ per image)
| Model id | ~Cost | Best for |
|---|
bytedance-seed/seedream-4.5 | ~$0.03 | Default — best quality-per-dollar, gpt-image-2-class |
recraft/recraft-v4 | ~$0.04 | UI / design / logos / vector output |
google/gemini-2.5-flash-image | ~$0.039 | Photoreal + in-image text (cheaper direct, though) |
black-forest-labs/flux.2-pro | ~$0.05 | Arguably highest quality |
openai/gpt-5-image | higher | OpenAI quality via one key |
For reference, OpenAI gpt-image-2 high ≈ $0.17–0.19/img — Seedream is ~6× cheaper.
Pricing changes — verify live: curl "https://openrouter.ai/api/v1/models?output_modalities=image".
How it works (different from OpenAI!)
OpenRouter generates images through the chat-completions endpoint with a modalities
param, NOT /images/generations:
POST https://openrouter.ai/api/v1/chat/completions
{ "model": "...", "messages": [{ "role": "user", "content": "PROMPT" }],
"modalities": ["image"], // ["image","text"] for gemini/openai models
"image_config": { "aspect_ratio": "16:9", "image_size": "2K" } }
Images return as base64 data URLs in choices[0].message.images[].image_url.url.
The helper handles modalities selection, parsing, and saving automatically.
Generate
set -a && source .env.local && set +a && npx tsx scripts/generate-image-openrouter.ts \
--prompt "PASTE PROMPT HERE" \
--model bytedance-seed/seedream-4.5 \
--aspect 16:9 \
--out designs/circletel-hero.png
Other models: --model black-forest-labs/flux.2-pro, --model recraft/recraft-v4.
Image-to-image / edit
set -a && source .env.local && set +a && npx tsx scripts/generate-image-openrouter.ts \
--edit designs/source.png \
--prompt "replace the background with a deep navy studio gradient" \
--out designs/edited.png
Aspect ratios
1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 9:16, 16:9, 21:9 (model-dependent).
gemini-3.1-flash-image-preview also supports 1:4, 4:1, 1:8, 8:1.
CircleTel brand context
Orange #F5831F, navy #1B2A4A, white accent. The prompt templates in
.claude/skills/nb-product-hero/SKILL.md are provider-agnostic — reuse them and append
colour-accurate orange #F5831F (NOT yellow or red), deep navy #1B2A4A.
Provider cheat-sheet
| Want | Use |
|---|
| Cheapest good quality | this skill → seedream-4.5 |
| Best instruction-following / transparent logos | openai-imagegen (gpt-image) |
| Photoreal + reliable in-image text | nb-* (Gemini direct) |
| Vector / icon / logo files | this skill → recraft/recraft-v4 |
