Skip to main content
The OpenAI Responses provider exposes an image generation tool that allows you to generate images on the provider.

​
Enable Image Generation

The image generation tool can be enabled via the chatModelOptions parameter of the Agent constructor: 
final agent = Agent(
  'openai-responses',
  chatModelOptions: const OpenAIResponsesChatModelOptions(
    serverSideTools: {OpenAIServerSideTool.imageGeneration},
    imageGenerationConfig: ImageGenerationConfig(
      partialImages: 3,
      quality: ImageGenerationQuality.low,
      size: ImageGenerationSize.square256,
    ),
  ),
);
ImageGenerationConfig lets you request progressive previews (partialImages), choose the quality profile, and select a size. Omit these values to fall back to the provider defaults.

​
Streaming Previews

Partial renders arrive as base64-encoded bytes inside metadata events: 
void handleMetadata(Map<String, dynamic> metadata) {
  final events = metadata['image_generation'] as List?;
  if (events == null) return;

  for (final event in events) {
    final preview = event['partial_image_b64'] as String?;
    if (preview != null) {
      // import 'dart:convert' to use base64Decode
      final bytes = base64Decode(preview);
      savePreview(bytes, event['partial_image_index'] as int);
    }
  }
}
Each event also includes the tool stage (response.image_generation_call.*) so you can display progress indicators while the image renders.

​
Final Image

Completed images are attached to the assistant message as DataPart instances with an image/* MIME type: 
final result = await agent.send(prompt);
for (final msg in result.messages) {
  for (final part in msg.parts) {
    if (part is DataPart && part.mimeType.startsWith('image/')) {
      saveFinalImage(part.bytes);
    }
  }
}
No extra download step is required; Dartantic fetches the file from the provider and hands you the bytes.

​
Example