Overview
Dust agents can generate images from text prompts using the Image Generation capability. This feature allows your agents to create custom visuals based on natural language descriptions, enabling a wide range of creative and practical applications. What sets Dust’s image generation apart is the ability to use reference images and run parallel generations. Agents can use previously generated images (or uploaded images) as references for subsequent generations, maintaining visual consistency across a series of assets. Combined with parallel execution, this enables complex asset creation pipelines where multiple variations are generated simultaneously. The Image Generation capability uses the gemini-3-pro-image model to transform your text descriptions into high-quality images. Whether you need product photography pipelines, brand asset variations, multi-format marketing campaigns, or visual style transfers, your agents can create them on demand.How to Enable Image Generation
When creating or editing an agent:- In the Agent Builder, click on Add Tools
- Select the Image Generation capability
- Save your agent
Tool Capabilities
The Image Generation tool provides several configurable options displayed in the UI:Quality Settings
Choose between quality levels based on your needs. Higher quality produces more detailed images but takes longer to generate.Full quality downloads: For performance reasons, image previews in the
conversation are compressed. Click the download button to get the generated
image at maximum quality.
Aspect Ratio Options
Select from multiple aspect ratios to fit your use case:Reference Images
Use images from previous tool calls or uploaded images as references for new generations. This maintains visual consistency across multiple outputs—essential for creating cohesive brand assets, product variations, or campaign materials.Currently up to 14 references per image generation.
Parallel Execution
Agents can run multiple image generations simultaneously. This enables efficient asset pipelines where several variations are created at once, dramatically speeding up workflows that require multiple related images.How to Use Image Generation
As a tool in an agent
When the Image Generation capability is enabled, your agent can autonomously decide to generate images when it determines they would be helpful for the task at hand. The agent evaluates your request and uses image generation when it’s appropriate to create visual content. For example, if you ask an agent with Image Generation enabled to “help me design a logo concept,” the agent will automatically use the image generation tool to create visual proposals.In conversations
You can directly ask any agent with Image Generation enabled to create images in your conversations. Simply describe what you want to see, and the agent will generate the image for you.Safety features
- All generated images are watermarked with SynthID
- NSFW and images including celebrities won’t be accepted
Advanced Usage
Agents can run multiple image generations in parallel and use reference images from previous outputs. This enables powerful creative pipelines that maintain consistency while producing multiple assets efficiently.E-commerce: Product Photography Pipeline
Real Estate: Property Staging
Best Practices
- Be descriptive: Provide clear, detailed descriptions of what you want to see
- Specify style: Mention artistic styles, color schemes, or reference styles when relevant
- Iterate: Don’t hesitate to ask for modifications if the first result isn’t quite right
- Context matters: Explain the purpose of the image to help the agent make better creative decisions
- Use references strategically: When maintaining consistency matters, explicitly tell the agent which previous image to use as reference
- Leverage parallel generation: For pipelines requiring multiple assets, describe all variations in a single request to enable parallel execution