Generate Image

Generate images using the Gemini per-call image model.

Supports text-to-image, image-to-image, and multi-image blending, all using the OpenAI Images-compatible request format.

Use Cases

Want to call Gemini image capabilities directly using public model names

Don't want to construct Gemini native contents / parts manually

Want to receive URL results uniformly for easy frontend or business-side display

Model Selection Guide

nano-banana: Basic version, fixed 1K

nano-banana-pro-*: Resolution (1K / 2K / 4K) distinguished by model name

nano-banana-2-*: Resolution (0.5K / 1K / 2K / 4K) distinguished by model name

Key Field Descriptions

`model`

The model field determines the underlying image capability, output resolution tier, and billing tier simultaneously.

For standard quality at the lowest cost: start with nano-banana

For higher quality while maintaining standard aspect ratios: use nano-banana-pro-*

For ultra-wide formats or Gen 2 capabilities: use nano-banana-2-*

Resolution is encoded in the model name, so no additional field is needed to select the tier.

`image`

Not provided: text-to-image

Single string: single-image image-to-image

String array: multi-image blending

It is recommended to pass publicly accessible image URLs.

If your client uses Base64, wrap it as a data:image/png;base64,... string yourself; the platform will not automatically convert raw Base64 to an image reference format.

`aspect_ratio`

This is a top-level field, not nested inside another object.

When omitted, the platform does not inject a default aspect ratio; the actual default follows the upstream model

nano-banana / nano-banana-pro-* are suitable for standard poster, landscape, and portrait ratios

nano-banana-2-* additionally supports ultra-wide formats like 1:4, 4:1, 1:8, 8:1

Adjusting aspect_ratio only changes the canvas ratio, not the model tier you selected

Response Description

Response structure follows the OpenAI Images style

data[].url is the final accessible image URL

data[].revised_prompt is the normalized prompt from the platform

application/json

Image generated successfully

Body

示例

Single image responseMulti-image blending response

{
    "created": 1773294301,
    "data": [
        {
            "url": "https://cdn.ccapi.ai/images/generated/nano-banana/1773294301681-0.png?token=example",
            "revised_prompt": "According my description, output image: A tiny banana astronaut sticker on a white background, flat vector style"
        }
    ]
}

Generate Image

Use Cases

Model Selection Guide

Key Field Descriptions

`model`

`image`

`aspect_ratio`

Response Description

请求参数

请求示例代码

返回响应

Generate Image

Use Cases#

Model Selection Guide#

Key Field Descriptions#

model#

image#

aspect_ratio#

Response Description#

请求参数

请求示例代码

返回响应

Use Cases

Model Selection Guide

Key Field Descriptions

`model`

`image`

`aspect_ratio`

Response Description