🇬🇧 EN

🎨 Image Generation

Generates an image from a text prompt using an AI image model such as DALL-E 3. The node returns a URL pointing to the generated image, along with the final prompt that was used. Wire this into any workflow that needs to create visual content dynamically.

Category: Data Retrieval & Processing · Type identifier: image

Overview

The Image Generation node connects to an AI image model via your configured AI Provider and generates an image based on a text description you supply. Because the prompt field supports variable references, you can build the description dynamically from earlier workflow steps — for example, asking an Agent node to write a detailed image prompt, then passing that prompt to this node.

Typical use cases include:

Generating product concept images from marketing briefs.
Creating personalised social media graphics based on user-provided descriptions.
Producing illustrations for automatically generated articles or reports.
Generating avatar images for new user accounts.
Creating visual mockups or diagrams from textual specifications.

Temporary URLs. Most image generation providers (including DALL-E 3) return a URL that expires after approximately one hour. If you need to keep the image permanently, download it immediately in the same workflow using an HTTP Request node and save it to file storage (Local File, Amazon S3, Google Drive, etc.).

Configuration

Field	Status	Description
Provider	Required	Select an AI Provider that supports image generation. The provider must be configured with a model capable of generating images. Not all text-generation providers support image generation — check your provider's capabilities.
Prompt	Required	A text description of the image to generate. Supports `{{ variable }}` references so you can build the prompt dynamically from earlier steps. Be as specific as possible — include subject, style, mood, colour palette, and composition for best results.
Size	Optional	The dimensions of the generated image. Options: `1024x1024` (square), `1024x1792` (portrait / tall), `1792x1024` (landscape / wide). Default is `1024x1024`.
Quality	Optional	`standard` generates faster at lower cost. `hd` produces higher detail and finer textures at higher cost. Currently only supported by DALL-E 3. Default is `standard`.
Style	Optional	`vivid` produces hyper-real, dramatic images with bold colours. `natural` produces more realistic, subdued images. Currently only supported by DALL-E 3. Default is `vivid`.

Output Data

Field	Type	Description
`url`	string	The URL of the generated image. This URL is temporary — save the image to permanent storage if you need to keep it.
`revised_prompt`	string	The prompt that was actually used by the model after any automatic revisions. The model may rephrase or expand your original prompt to improve image quality. Review this field if the output does not match your expectations.
`size`	string	The dimensions of the image that was generated (e.g. `1024x1024`).

// Use the generated image URL in a later step {{ generate_image.output.url }} // See what prompt was actually used {{ generate_image.output.revised_prompt }} // Reference in an email body (HTML) <img src="{{ generate_image.output.url }}" alt="Generated image" />

Example Usage

Auto-generating a social media image from an article headline

Receive the article headline via webhook. The trigger payload contains headline and topic.
Use an Agent node to write a detailed image prompt. Prompt the agent:
Write a detailed image generation prompt for a social media post about the following article headline. The image should be photorealistic, modern, and suitable for a professional technology blog. Headline: {{ trigger.output.headline }} Topic: {{ trigger.output.topic }} Return only the image prompt, nothing else.
Add an Image Generation node. Set Provider to your image-generation-capable provider. Set Prompt to {{ write_prompt.output.text }}. Set Size to 1792x1024 (landscape, ideal for social media headers). Set Quality to hd.
Save the image to permanent storage. Add an HTTP Request node (GET {{ generate_image.output.url }}) to download the image bytes, then pass them to an Amazon S3 or Local File node for permanent storage. Reference the saved file path in subsequent steps.

Tips & Notes

Save generated images immediately. DALL-E 3 URLs expire after approximately one hour. If your workflow passes the URL to another step that runs much later, or if you store the URL in a database, the image may be gone by the time it is accessed. Download and store it in the same run.
Prompt quality has a huge impact on output quality. Short, vague prompts produce generic results. Use the Agent node upstream to write a rich, detailed prompt — describing subject, style, lighting, composition, colour, and mood — before passing it to this node.
Use revised_prompt for debugging. If the generated image does not look right, check revised_prompt to see how the model interpreted your input. The model sometimes modifies prompts in ways that change the output.
Quality and Style are provider-specific. These options currently only apply to DALL-E 3 via OpenAI. If you are using a different provider, these settings may be silently ignored. Check your provider's documentation.
Content policies apply. Image generation models enforce content policies. Prompts requesting harmful, violent, explicit, or copyright-infringing content will be rejected by the provider. Flusso will mark the step as failed with the provider's error message.

Related Nodes

Agent — use an agent to write a high-quality image generation prompt before passing it to this node.