LLM Image

Turn a text prompt into a generated image URL for your workflow result.

Audience

Creators building and publishing workflows.

Overview#

LLM Image creates visuals from a prompt using Gemini or OpenAI image models. Default is Nano Banana 2 (Gemini 3.1 Flash Image).

It works best after upstream steps have already shaped a clear, specific prompt, not vague raw customer input.

Input: Prompt: The image description, usually polished text from an upstream LLM or input block.
Output: Image URL: A link to the generated image for display or downstream use.

Prompt: Fallback description when Prompt input is not connected.
Model: Image provider and generation behavior.
Aspect ratio: Frame shape. Gemini uses this directly; OpenAI maps to the nearest supported shape.
Quality: Low, medium, or high. This applies to OpenAI image models only and is ignored for Gemini.

Refine the prompt upstream before image generation; this block generates the image.
Pick aspect ratio to match the product (square, landscape, portrait).
Treat image generation as a late-stage step in the graph.
Confirm the run modal shows images correctly when Image URL feeds Workflow Output.

Was this useful?

Your response helps us improve the documentation.

Ask the customer one clear question and turn their answer into data your workflow can use.

Collect several run-modal fields in one grouped form, each with a stable key for downstream steps.

Define what the customer receives when the workflow finishes.

Turn a text prompt into a generated image URL for your workflow result.

Audience

Creators building and publishing workflows.

LLM Image creates visuals from a prompt using Gemini or OpenAI image models. Default is Nano Banana 2 (Gemini 3.1 Flash Image).

It works best after upstream steps have already shaped a clear, specific prompt, not vague raw customer input.

Input: Prompt: The image description, usually polished text from an upstream LLM or input block.
Output: Image URL: A link to the generated image for display or downstream use.

Prompt: Fallback description when Prompt input is not connected.
Model: Image provider and generation behavior.
Aspect ratio: Frame shape. Gemini uses this directly; OpenAI maps to the nearest supported shape.
Quality: Low, medium, or high. This applies to OpenAI image models only and is ignored for Gemini.

Refine the prompt upstream before image generation; this block generates the image.
Pick aspect ratio to match the product (square, landscape, portrait).
Treat image generation as a late-stage step in the graph.
Confirm the run modal shows images correctly when Image URL feeds Workflow Output.