API documentation

Gemini Images

Generate images with Gemini image models through chat completions.

Gemini image models work as chat models that can return images. Use /v1/chat/completions for both generation and editing, not /v1/images/generations and not /v1/images/edits. Supported ids include google/gemini-3.1-flash-image-preview and google/gemini-3-pro-image; use the exact model id from the catalog or /v1/models.

Request

Gemini image request

curl https://api.aigate.shop/v1/chat/completions \  -H "Authorization: Bearer sk-your-api-key" \  -H "Content-Type: application/json" \  -d '{    "model": "google/gemini-3.1-flash-image-preview",    "messages": [      {        "role": "user",        "content": "Generate an image of a small yellow banana-shaped robot on a white desk, studio lighting"      }    ],    "modalities": ["image", "text"]  }'

Image config

For aspect ratio and output size, send image_config at the top level of the request. aspect_ratio controls the aspect ratio, and image_size controls output size/quality such as 1K, 2K, or 4K when the selected model supports it.

Thunder Client JSON body

{  "model": "google/gemini-3-pro-image",  "messages": [    {      "role": "user",      "content": "Generate a sharp futuristic product render of a matte black AI gateway device on a white desk, studio lighting"    }  ],  "modalities": ["image", "text"],  "image_config": {    "aspect_ratio": "16:9",    "image_size": "2K"  }}

Response

Gemini image response

{  "id": "gate-1779703910",  "object": "chat.completion",  "created": 1779703910,  "model": "google/gemini-3.1-flash-image-preview",  "choices": [    {      "index": 0,      "message": {        "role": "assistant",        "content": null,        "images": [          {            "type": "image_url",            "image_url": {              "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA..."            }          }        ]      },      "finish_reason": "stop"    }  ]}

Edit an existing image

For editing, pass the source image as image_url inside messages. The text in the same message describes what to change. The model returns a new image as base64 inside a data URL.

Gemini image edit request

curl https://api.aigate.shop/v1/chat/completions \  -H "Authorization: Bearer sk-your-api-key" \  -H "Content-Type: application/json" \  -d '{    "model": "google/gemini-3.1-flash-image-preview",    "messages": [      {        "role": "user",        "content": [          {            "type": "text",            "text": "Edit this image: add a small orange cat sitting on the table. Keep the original lighting and style. Return the edited image."          },          {            "type": "image_url",            "image_url": {              "url": "https://picsum.photos/seed/aigate-edit-test/800/600"            }          }        ]      }    ],    "modalities": ["image", "text"]  }'

Gemini image edit response

{  "id": "gate-1779703910",  "object": "chat.completion",  "created": 1779703910,  "model": "google/gemini-3.1-flash-image-preview",  "choices": [    {      "index": 0,      "message": {        "role": "assistant",        "content": null,        "images": [          {            "type": "image_url",            "image_url": {              "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA..."            }          }        ]      },      "finish_reason": "stop"    }  ]}

Where to read the image

txt

choices[0].message.images[0].image_url.url

The response is a base64 data URL

image_url.url usually looks like data:image/png;base64,... . To save a file, take the part after the comma and decode base64. content can be null for image output.

Request fields

Field	Required	Meaning
model	yes	Exact model id from the catalog.
messages	yes	Put the prompt in a user message.
messages.content[].image_url	edits/input image	Source image for editing or image input.
modalities	yes	["image", "text"]
image_config	no	Gemini image settings: aspect_ratio (for example 1:1 or 16:9) and image_size (for example 1K, 2K, or 4K when supported by the model).

Save a data URL

import { writeFileSync } from "node:fs";const url = response.choices[0].message.images[0].image_url.url;if (url.startsWith("data:image/")) {  const [, base64] = url.split(",");  writeFileSync("gemini-image.png", Buffer.from(base64, "base64"));}

SillyImages

For the SillyImages SillyTavern extension, choose the Gemini API type. Put the base URL without /v1 in the endpoint field: https://api.aigate.shop. The extension sends generation to /v1beta/models/{model}:generateContent and loads models from /v1/models by itself.

SillyImages settings

API type: Gemini (nano-banana)Endpoint URL: https://api.aigate.shopAPI key: sk-your-api-keyModel: google/gemini-3-pro-imageAspect ratio: 16:9Image size: 2K

The compatible route accepts the Google contents/generationConfig shape and returns candidates/content/parts/inlineData so the extension can read the image directly.

Google-compatible request

curl https://api.aigate.shop/v1beta/models/google/gemini-3-pro-image:generateContent \  -H "Authorization: Bearer sk-your-api-key" \  -H "Content-Type: application/json" \  -d '{    "contents": [      {        "role": "user",        "parts": [          {            "text": "Generate a cinematic 16:9 image of a tiny banana-shaped robot in a rainy neon street"          }        ]      }    ],    "generationConfig": {      "responseModalities": ["TEXT", "IMAGE"],      "imageConfig": {        "aspectRatio": "16:9",        "imageSize": "2K"      }    }  }'

Google-compatible response

{  "candidates": [    {      "content": {        "role": "model",        "parts": [          {            "inlineData": {              "mimeType": "image/png",              "data": "iVBORw0KGgoAAAANSUhEUgAA..."            }          }        ]      },      "finishReason": "STOP",      "index": 0    }  ],  "usageMetadata": {    "promptTokenCount": 18,    "candidatesTokenCount": 1290,    "totalTokenCount": 1308  }}

Common mistakes

Do not send Gemini image models to /v1/images/generations or /v1/images/edits.
Do not forget modalities: ["image", "text"].
For size and aspect ratio, use image_config at the top level of the request.
Do not read only message.content.
Check the exact model id in /v1/models or the catalog.

GPT Image

Video Generation