Image Generation
Generate images from text prompts using AI. Your agent can create illustrations, diagrams, logos, and more — all saved automatically to the Asset Library. Nano Banana 2 is the default model; Imagen 4 is deprecated and shuts down August 17, 2026.
Setup
Enable Image Generation in your agent's Settings. Then add a provider credential:
| Provider | Credential | Notes |
|---|---|---|
| Gemini / Nano Banana | Gemini API key | Default Nano Banana 2 plus Nano Banana Pro, supports image input and edits |
| OpenAI | OpenAI API key | Direct GPT Image 2 text-to-image; separately metered API access, not a ChatGPT subscription |
| OpenRouter | OpenRouter API key | Curated image models with optional reference-image input |
| Nano Banana | Nano Banana API key | Multiple model tiers (generate, generate-2, generate-pro) |
You can select your preferred provider and model in the chat interface using the model selector.
Tools
| Tool | Description |
|---|---|
| generate_image | Generate an image from a text prompt and save it to assets |
| list_assets | List all generated images and files in the asset library |
Parameters
generate_image
| Parameter | Type | Required |
|---|---|---|
| prompt | string | Yes |
| name | string | Yes |
| provider | "gemini" | "openai" | "openrouter" | "nano_banana" | No |
| width | number | No (default: 1024) |
| height | number | No (default: 1024) |
| folder_id | string | No |
| input_asset_id | string | No; only used by models with image input support |
Image Models
| Model | Status | Notes |
|---|---|---|
| gemini:gemini-3.1-flash-image | Default | Stable Nano Banana 2, supports image input and edits |
| gemini:gemini-3.1-flash-lite-image | Available | Stable Nano Banana 2 Lite for low-latency 1K generation and editing |
| gemini:gemini-3-pro-image | Available | Stable Nano Banana Pro for higher-fidelity image work |
| openai:gpt-image-2 | Available | Direct OpenAI text-to-image; reference input is disabled in this flow |
| openrouter:openai/gpt-image-2 | Available | GPT Image 2 through OpenRouter with reference-image input |
| openrouter:bytedance-seed/seedream-4.5 | Available | Seedream 4.5 through OpenRouter |
| openrouter:black-forest-labs/flux.2-pro | Available | FLUX.2 Pro through OpenRouter |
| openrouter:recraft/recraft-v4.1 | Available | Recraft V4.1 through OpenRouter |
| nano_banana:generate-2 | Available | Third-party Nano Banana API model with custom resolution |
| nano_banana:generate-pro | Available | Third-party Nano Banana API high-resolution model |
| gemini:imagen-4.0-generate-001 | Deprecated | Imagen 4 shuts down August 17, 2026 |
| gemini:imagen-4.0-fast-generate-001 | Deprecated | Imagen 4 Fast shuts down August 17, 2026 |
| gemini:imagen-4.0-ultra-generate-001 | Deprecated | Imagen 4 Ultra shuts down August 17, 2026 |
| nano_banana:generate | Legacy | Older Nano Banana API model |
Supported Aspect Ratios
The selected aspect ratio is forwarded to native Gemini models and to curated OpenRouter models that expose aspect-ratio control. Direct GPT Image 2 uses the nearest valid pixel dimensions.
Example
User: "Generate a logo for my coffee shop called Bean There"
Agent: Generates image → saves to assets → returns preview with download link
Integration
Generated images are stored in the Asset Library with full metadata (prompt, provider, model, dimensions). Combine with Automations for workflows like "when a task is created → generate a thumbnail."