---
title: "Gemini Image Generation Migration Guide"
description: "Anyone using the following models with /chat/completions: - `gemini/gemini-2.0-flash-exp-image-generation` - `vertex_ai/gemini-2.0-flash-exp-image-generation`"
type: skill
canonical_url: https://claudary.paisolsolutions.com/skills/gemini-img-migration
source: "Claudary"
difficulty: intermediate
author: "Claude Code Knowledge Pack"
date: 2026-07-10T11:24:45.165Z
license: CC-BY-4.0
attribution: "Gemini Image Generation Migration Guide — Claudary (https://claudary.paisolsolutions.com/skills/gemini-img-migration)"
---

# Gemini Image Generation Migration Guide
Anyone using the following models with /chat/completions: - `gemini/gemini-2.0-flash-exp-image-generation` - `vertex_ai/gemini-2.0-flash-exp-image-generation`

## Overview

# Gemini Image Generation Migration Guide

## Who is impacted by this change?

Anyone using the following models with /chat/completions:
- `gemini/gemini-2.0-flash-exp-image-generation`
- `vertex_ai/gemini-2.0-flash-exp-image-generation`

## Key Change

:::info
From v1.77.0, LiteLLM will return the List of images in `response.choices[0].message.images` instead of a single image in `response.choices[0].message.image`.
:::

Gemini models now support image generation through chat completions. Images are returned in `response.choices[0].message.images` with base64 data URLs.

## Before and After

### Before
```python
from litellm import completion

response = completion(
    model="gemini/gemini-2.0-flash-exp-image-generation",
    messages=[{"role": "user", "content": "Generate an image of a cat"}],
    modalities=["image", "text"],
)


base_64_image_data = response.choices[0].message.content
```

### After  
```python
from litellm import completion

response = completion(
    model="gemini/gemini-2.0-flash-exp-image-generation",
    messages=[{"role": "user", "content": "Generate an image of a cat"}],
    modalities=["image", "text"],
)

# Image is now available in the response
image_url = response.choices[0].message.images[0]["image_url"]["url"]  # "data:image/png;base64,..."
```

### Why the change?

Because the newer `gemini-2.5-flash-image-preview` model sends both text and image responses in the same response. This interface allows a developer to explicitly access the image or text components of the response. Before a developer would have needed to search through the message content to find the image generated by the model.

**Why the change from `image` to `images`?**
This is to be consistent with the OpenRouter API, making sure we are using simple, well-known interfaces where possible.

## Usage

### Using the Python SDK

**Key Change:**
```diff
# Before
-- base_64_image_data = response.choices[0].message.content

# After
++ image_url = response.choices[0].message.images[0]["image_url"]["url"]
```

#### Basic Image Generation

```python
from litellm import completion
import os

# Set your API key
os.environ["GEMINI_API_KEY"] = "your-api-key"

# Generate an image
response = completion(
    model="gemini/gemini-2.0-flash-exp-image-generation",
    messages=[{"role": "user", "content": "Generate an image of a cat"}],
    modalities=["image", "text"],
)

# Access the generated image
print(response.choices[0].message.content)  # Text response (if any)
print(response.choices[0].message.images[0])    # Image data
```

#### Response Format

The image is returned in the `message.images` field:

```python
{
    "image_url": {
        "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA...",
        "detail": "auto"
    },
    "index": 0,
    "type": "image_url"
}
```

### Using the LiteLLM Proxy Server

**Key Change:**
```diff
# Before
-- "content": "base64-image-data..."

# After  
++ "images": [{
++   "image_url": {
++     "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA...",
++     "detail": "auto"
++   },
++   "index": 0,
++   "type": "image_url"
++ }]
```

#### Configuration Setup

1. **Configure your models in `config.yaml`:**

```yaml
model_list:
  - model_name: gemini-image-gen
    litellm_params:
      model: gemini/gemini-2.0-flash-exp-image-generation
      api_key: os.environ/GEMINI_API_KEY
  - model_name: vertex-image-gen  
    litellm_params:
      model: vertex_ai/gemini-2.5-flash-image-preview
      vertex_project: your-project-id
      vertex_location: us-central1

general_settings:
  master_key: sk-1234  # Your proxy API key
```

2. **Start the proxy server:**

```bash
litellm --config /path/to/config.yaml

# RUNNING on http://0.0.0.0:4000
```

#### Making Requests

**Using OpenAI SDK:**

```python
from openai import OpenAI

# Point to your proxy server
client = OpenAI(
    api_key="sk-1234",  # Your proxy API key
    base_url="http://0.0.0.0:4000"
)

response = client.chat.completions.create(
    model="gemini-image-gen",
    messages=[{"role": "user", "content": "Generate an image of a cat"}],
    extra_body={"modalities": ["image", "text"]}
)

# Access the generated image
print(response.choices[0].message.content)  # Text response (if any)
print(response.choices[0].message.image)    # Image data
```

**Using curl:**

```bash
curl -X POST 'http://0.0.0.0:4000/v1/chat/completions' \\
-H 'Content-Type: application/json' \\
-H 'Authorization: Bearer sk-1234' \\
-d '{
  "model": "gemini-image-gen",
  "messages": [
    {
      "role": "user",
      "content": "Generate an image of a cat"
    }
  ],
  "modalities": ["image", "text"]
}'
```

**Response format from proxy:**

```json
{
  "id": "chatcmpl-123",
  "object": "chat.completion",
  "created": 1704089632,
  "model": "gemini-image-gen",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here's an image of a cat for you!",
        "images": [{
          "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA...",
          "detail": "auto"
        }
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 8,
    "total_tokens": 18
  }
}
```

---

Source: [Claudary](https://claudary.paisolsolutions.com/skills/gemini-img-migration) · https://claudary.paisolsolutions.com