Image generation

POST /v1/images/generations

Drop-in compatible with the OpenAI images.generate SDK. Provider-specific parameters live alongside the standard ones — the backend forwards what each model understands.

Common parameters

Basic example

curl https://api.airforce/v1/images/generations \
  -H "Authorization: Bearer sk-air-YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-2",
    "prompt": "A cute baby sea otter floating on its back",
    "n": 1,
    "response_format": "url"
  }'

Response shape

{
  "created": 1715000000,
  "data": [{
    "url": "https://.../image.jpeg"
  }]
}

Streaming variant

Pass sse: true when proxying through middleware that has tight idle-timeouts. The response is one SSE event with the complete JSON, then [DONE].

data: {"created":1715000000,"data":[{"url":"https://.../image.jpeg"}]}

data: [DONE]

Image-to-image & reference

Models with image_caps.max_input_images > 0 accept reference images. Either pass URLs or base64-encoded data.

curl https://api.airforce/v1/images/generations \
  -H "Authorization: Bearer sk-air-YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-2",
    "prompt": "Same character, now in a snowy forest",
    "aspect_ratio": "16:9",
    "input_images": [
      {"url": "https://example.com/character.jpg"}
    ]
  }'

Image edits

The OpenAI-compatible edit endpoint. Send the source image as a multipart file, which is what the OpenAI SDKs do for images.edit, or send the JSON body shown above with input_images. It runs the same path as /v1/images/generations: the same models, prices and response shape.

curl https://api.airforce/v1/images/edits \
  -H "Authorization: Bearer sk-air-YOUR_API_KEY" \
  -F "model=nano-banana-2" \
  -F "prompt=Same character, now in a snowy forest" \
  -F "[email protected]" \
  -F "aspect_ratio=16:9"

from openai import OpenAI

client = OpenAI(base_url="https://api.airforce/v1", api_key="sk-air-YOUR_API_KEY")

with open("character.png", "rb") as image:
    result = client.images.edit(
        model="nano-banana-2",
        image=image,
        prompt="Same character, now in a snowy forest",
    )
print(result.data[0].url)

Up to 16 input images per request, 25 MB each; the model’s own limit is image_caps.max_input_images. A request without an input image is refused with 400 instead of becoming a text-to-image call. Each input image can add to the price, as shown on the model page.

Model-specific parameters

Beyond the common parameters above, individual model families accept extras. Unknown parameters are forwarded to the upstream as-is, so any future provider-specific knob also works without an SDK update.

Nano-Banana

nano-banana-2, nano-banana-pro

Suno (music)

suno-v4.5, suno-v5 and suno-v5.5 were historically accepted on this endpoint. They now belong to the native music API at /v1/audio/music — calls here are rejected with `wrong_endpoint`.

Discovering capabilities

Each image model exposes image_caps on /v1/models so clients can validate inputs before sending:

Image pricing is not on image_caps — use the model’s price fields on /v1/models for cost.