ContextQMD
Back to Eliza

Image Generation Plugin

packages/docs/plugin-registry/image-generation.md

2.0.14.1 KB
Original Source
<Warning> This plugin is not yet available in the Eliza plugin registry. For image generation, use providers that support it natively (OpenAI DALL-E, Google Imagen) through their respective provider plugins. </Warning>

The Image Generation plugin enables Eliza agents to generate images using AI image models, including OpenAI's DALL-E, Stable Diffusion via FAL, and other providers.

On-demand plugin. This plugin is resolved from the remote elizaOS plugin registry and auto-installs when configured. It is not included in Eliza's bundled plugins.json index.

Package: @elizaos/plugin-image-generation

Overview

The Image Generation plugin registers an IMAGE model handler and a set of actions that allow agents to generate images from text descriptions, edit existing images, and create variations.

Installation

bash
eliza plugins install @elizaos/plugin-image-generation

Enable via Features

json
{
  "features": {
    "imageGen": true
  }
}

Configuration

The plugin supports multiple image generation backends. Configure via the media.image section:

json
{
  "media": {
    "image": {
      "enabled": true,
      "mode": "own-key",
      "provider": "openai",
      "model": "dall-e-3"
    }
  }
}

OpenAI DALL-E

Environment VariableRequiredDescription
OPENAI_API_KEYYesOpenAI API key
json
{
  "media": {
    "image": {
      "provider": "openai",
      "model": "dall-e-3",
      "size": "1024x1024",
      "quality": "standard"
    }
  }
}

FAL (Stable Diffusion and others)

Environment VariableRequiredDescription
FAL_API_KEYYesFAL API key from fal.ai
json
{
  "media": {
    "image": {
      "provider": "fal",
      "model": "fal-ai/flux/schnell"
    }
  }
}

Supported Providers and Models

OpenAI

ModelDescription
dall-e-3Highest quality, 1024×1024 to 1792×1024
dall-e-2Legacy, 256×256 to 1024×1024

FAL

Model IDDescription
fal-ai/flux/schnellFLUX Schnell — fast, high quality
fal-ai/flux/devFLUX Dev — more detailed
fal-ai/flux-proFLUX Pro — best quality
fal-ai/stable-diffusion-v3-mediumSD3 Medium
fal-ai/stable-video-diffusionVideo generation

Actions

ActionDescription
GENERATE_MEDIAGenerate an image from a text description using mediaType: "image"
EDIT_IMAGEEdit an existing image with a prompt
IMAGE_VARIATIONCreate variations of an existing image
DESCRIBE_IMAGEAnalyze and describe an image (uses vision)

Usage Examples

After the plugin is loaded:

"Draw a cozy coffee shop on a rainy afternoon, watercolor style"

"Generate an image of a robot reading a book in a library"

"Create a logo for an AI company called Eliza"

The agent generates the image and can share it in the conversation or save it to the workspace.

Output Handling

Generated images are:

  • Returned as URLs (hosted temporarily by the provider)
  • Optionally downloaded and saved to the agent workspace
  • Embeddable in supported platform connectors (Discord, Telegram)

Size and Quality Options

DALL-E 3 Sizes

SizeAspect Ratio
1024x1024Square (default)
1792x1024Landscape
1024x1792Portrait

DALL-E 3 Quality

QualityDescription
standardFaster, lower cost
hdHigher detail, slower, higher cost

Media Configuration

For FAL auto-enable, set in media.image:

json
{
  "media": {
    "image": {
      "enabled": true,
      "mode": "own-key",
      "provider": "fal"
    }
  }
}

This triggers the FAL plugin to load automatically.