---
license: mit
tags:
- endpoints-template
- gemini
- image-editing
- gemini-2.0-flash-exp
library_name: google-genai
---

# Gemini Image Edit Model

This model uses Google's Gemini 2.0 Flash Experimental API to edit images based on text prompts.

## Usage

### Input Format
```json
{
  "inputs": {
    "image": "base64_encoded_image",
    "prompt": "Your editing instruction",
    "gemini_api_key": "your_gemini_api_key"
  }
}
```

### Output Format
When image is generated:
```json
{
  "generated_outputs": [
    {
      "image": {
        "url": "data:image/png;base64,edited_image_data",
        "mime_type": "image/png"
      }
    }
  ],
  "gemini_output": "",
  "prompt_used": "Your editing instruction"
}
```

When text is returned instead:
```json
{
  "generated_outputs": [],
  "gemini_output": "Text response from Gemini",
  "prompt_used": "Your editing instruction"
}
```

## Environment Variables

You can set `GEMINI_API_KEY` as an environment variable instead of passing it in each request.

## Model Details

- Uses Gemini 2.0 Flash Experimental model
- Supports image editing based on text prompts
- Returns either edited images or text responses
- Temperature: 1.0, Top-p: 0.95, Top-k: 40
- Max output tokens: 8192

## Custom Handler

This model uses a custom inference handler to integrate with Google's Gemini API for image editing tasks.