> ## Documentation Index
> Fetch the complete documentation index at: https://docs.emby.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Image Generation

> Generate images with Emby using the standard chat completions API — fully OpenAI-compatible.

# Image Generation

Emby supports image-generation models through the same **OpenAI-compatible `/v1/chat/completions` API** you already use for text.\
If your app already supports OpenAI image responses, it works with Emby automatically.

Generated images are returned as **base64-encoded data URLs** inside the assistant response.

***

## Available Models

You can view all image-generation capable models directly from:

```bash theme={null}
GET https://api.emby.dev/v1/models
```

Models such as **qwen2.5-vl**, **glm-vision**, **deepseek-vision**, and others support both generation and vision editing workflows.

***

## Generate an Image

Image generation works exactly like a normal chat request — just choose a vision-enabled or image-generating model.

```bash theme={null}
curl -X POST "https://api.emby.dev/v1/chat/completions" \
  -H "Authorization: Bearer $EMBY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen2.5-vision",
    "messages": [
      {
        "role": "user",
        "content": "Generate an image of a golden retriever puppy playing in a sunny meadow"
      }
    ]
  }'
```

***

## Response Format

Emby returns image results using the standard OpenAI-style structure:

```json theme={null}
{
  "id": "chatcmpl-123",
  "object": "chat.completion",
  "created": 1730000000,
  "model": "qwen2.5-vision",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here is your generated image:",
        "images": [
          {
            "type": "image_url",
            "image_url": {
              "url": "data:image/png;base64,<base64data>"
            }
          }
        ]
      },
      "finish_reason": "stop"
    }
  ]
}
```

The important part is the `images` array:

**Image Item Structure**

* `type`: always `"image_url"`
* `image_url.url`: base64-encoded `data:image/...` URL

***

## Vision Support (Editing & Modifying Images)

You can send an image **to the model** for editing, analysis, or transformation.\
Simply include a `type: "image_url"` or `type: "file"` object inside the messages.

Example:

```json theme={null}
{
  "role": "user",
  "content": [
    {
      "type": "image_url",
      "image_url": { "url": "data:image/png;base64,<your_image>" }
    },
    {
      "type": "text",
      "text": "Remove the background and convert this to a 3D style."
    }
  ]
}
```

Emby vision models can:

* Edit images
* Describe images
* Extract text
* Generate variations
* Combine text + image tasks

***

## Using the Emby AI SDK (Full Example)

Below is a clean, ready-to-copy example using the Emby AI SDK provider.

### `/api/chat/route.ts`

```ts theme={null}
import { streamText, convertToModelMessages } from "ai";
import { createEmby } from "@emby/ai-sdk-provider";

export async function POST(req: Request) {
  const body = await req.json();
  const emby = createEmby({
    apiKey: process.env.EMBY_API_KEY!,
    baseUrl: "https://api.emby.dev/v1"
  });

  try {
    return streamText({
      model: emby.chat("qwen2.5-vision"),
      messages: convertToModelMessages(body.messages)
    }).toUIMessageStreamResponse();
  } catch {
    return new Response(
      JSON.stringify({ error: "Emby request failed" }),
      { status: 500 }
    );
  }
}
```

***

## Rendering in Your Frontend

Below is a minimal, Emby-cleaned version of image rendering logic:

```ts theme={null}
import { extractBase64FromDataUrl } from "@/lib/image-utils";

export const ImageMessage = ({ image }) => {
  const base64 = extractBase64FromDataUrl(image.url);
  return (
    <img
      src={image.url}
      alt="Generated"
      className="rounded-lg shadow-md"
    />
  );
};
```

***

## Image Utils (Compatible With Emby)

```ts theme={null}
export function extractBase64FromDataUrl(dataUrl: string) {
  if (!dataUrl.startsWith("data:")) return "";
  return dataUrl.substring(dataUrl.indexOf(",") + 1);
}
```

***

## Usage Notes

<CardGroup cols={2}>
  <Card title="Larger Payloads" icon="file-image">
    Image responses are base64-encoded, so response sizes can be large.
  </Card>

  <Card title="Higher Compute Cost" icon="gauge">
    Image models typically use more compute than text-only models.
  </Card>

  <Card title="Streaming Support" icon="bolt">
    Emby can stream text; image parts arrive once the model finishes synthesis.
  </Card>

  <Card title="EU Hosting" icon="shield">
    All Emby-hosted vision models run inside EU data centers (bit.nl + Nebius AMS).
  </Card>
</CardGroup>

***

Need help integrating image generation?

**WhatsApp us:** [https://wa.absolum.nl](https://wa.absolum.nl)\
**Book a call:** [https://cal.com/absolum/30min](https://cal.com/absolum/30min)