فا

Images

Image generation — billed per image, capped at 10 per request.

POSThttps://iranrouter.com/v1/images/generations
Images always come back as b64_json, even when the model answers with a URL: the file is fetched once on our side and returned inline. response_format: url is accepted, but the answer is still b64_json.
Shell
curl https://iranrouter.com/v1/images/generations \
  -H "Authorization: Bearer ir-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-1",
    "prompt": "یک گربه روی پشت‌بام تهران، نقاشی آبرنگ",
    "n": 1,
    "size": "1024x1024"
  }'

How it is billed

  • The price is per image, so n is a direct cost multiplier. Newer models bill tokens as well; whichever the model quotes is what applies.
  • A hard ceiling of n = 10 per request — a larger value is clamped to 10 rather than rejected, and that same count is reserved from your wallet before the request goes out.
  • The image count on the usage row is the length of data in the response — what was actually produced.

For the cost breakdown, see billing.