Reference Images

Reference Image Nodes generate AI images from text prompts and optional source images. They are the primary node for creating images in Dal Nulla — from text-to-image generation to style transfer and image-to-image variations. Generated images can feed into Scene Nodes as visual references for video generation.

What is a Reference Image Node?

A Reference Image Node is the core building block for all image generation in Dal Nulla. It takes a text description (prompt) and optionally one or more source images, then produces an AI-generated image using your chosen model. The output image can be used standalone, connected to Scene Nodes for video generation, fed into Upscaler Nodes for enhancement, or referenced by other nodes using the tag system.

Inputs & Outputs

Inspector Controls

When you select a Reference Image Node, the Inspector Panel on the right provides the following controls:

Generation Prompt

A text area for describing the image to generate. If a Text Node is connected, its content appears here. You can click Detach to break the link and edit the prompt independently.

Model

Select the AI model for image generation. Each model has different strengths, resolution limits, and feature support. See the Available Models table below.

Aspect Ratio

Choose output dimensions: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3. Not all models support all ratios — if a ratio is rejected, try 1:1 or 16:9 as fallbacks.

Resolution

Output resolution, which varies by model. Most support up to 2K, while Seedream models go up to 4K. Higher resolutions consume more credits.

Source Images

Section showing connected source images (up to 3). Used for img2img, style transfer, and visual guidance. Connect images from other Reference, Draw, Upscaler, or Network Image nodes.

Batch Size

Number of images to generate in one run. Set a value greater than 1 to produce multiple variations, then lock the best result.

Tag

Assign a tag like @character-front to reference this image in other node prompts using the @tagName syntax. Tags must be unique within the project.

Use Global Context

Include project-wide context in generation. When enabled, global text and image references from the project will influence this generation.

Negative Prompt

Describe what to avoid in the generated image (e.g., “blur”, “text”, “distortion”). Not supported by all models — FLUX and Seedream models rely entirely on the positive prompt.

Steps / CFG / Seed

Advanced parameters for fine-tuning generation. Availability varies by model. Use a fixed seed for reproducible results.

Version History

Browse and restore previous generations. Every time you generate an image, the previous version is saved here.

Available Models

How to Use

1. Add a Reference Image Node to the canvas by right-clicking and selecting “Reference Image” from the menu, or by dragging it from the sidebar.
2. Write a description of the image in the prompt area, or connect a Text Node to provide the prompt.
3. Select an AI model based on your needs. Use free models (Nano Banana) for quick iterations and Pro models for final quality.
4. Choose the aspect ratio and resolution for your output image.
5. Optionally connect source images for img2img generation or style transfer by linking other Reference, Draw, or Upscaler nodes to the source port.
6. Assign a tag (e.g., @hero-portrait) so other nodes can reference this image in their prompts using @tagName syntax.
7. Click Generate or run the entire workflow to produce the image.

Workflow Example

Character Sheet Pipeline:

Create a Text Node with the prompt “a fantasy warrior with red armor, full body, concept art style” and connect it to three Reference Image Nodes, each with a variation (front view, side view, back view). Tag each one (@warrior-front, @warrior-side, @warrior-back). Then connect them to Scene Nodes that reference @warrior-front in their prompts — the AI will use the generated images as visual context for consistent character video.

Tips & Best Practices

• Connect up to 3 source images for stronger visual guidance in img2img mode.
• Use tags (@character, @style-ref) to reference generated images in Scene Node prompts — the AI will use them as visual context.
• Free models (Nano Banana, Nano Banana Pro) are great for quick iterations; switch to Pro models for final quality.
• For style transfer, connect a style reference image to the source port and describe the subject in the prompt.
• Batch generation: set Batch Size greater than 1 to generate multiple variations, then lock the best one.
• Models without negative prompt support (FLUX, Seedream) rely entirely on the positive prompt — be more descriptive about what you want.
• Mark important reference images as Global to automatically influence all scenes in the project.
• Use a fixed seed when you need reproducible results across multiple generations.

Troubleshooting

Blurry or Low-Quality Output

Check the resolution settings. Some models default to lower resolution. Try Seedream 4.5 for ultra-high resolution output, or use the Upscaler Node to enhance results after generation.

Style Drift Between Generations

When using source images for consistency, ensure the prompt describes the same style across all connected nodes. Use a specific model combined with a fixed seed for reproducibility. Tags help maintain visual consistency across the workflow.

Source Images Being Ignored

Not all models support img2img. For example, Grok Imagine does not accept source images. Check the “Img2Img” column in the model table above and switch to a model that supports it.

Aspect Ratio Rejected

Some models do not support all aspect ratios. If your chosen ratio causes an error, try 1:1 or 16:9 as fallbacks, which are universally supported.

Generation Taking Too Long

Higher resolutions and certain models (especially Seedream 4K) require significantly more processing time. Try reducing the resolution or switching to a faster model (Nano Banana, GPT Image 1 Mini, Riverflow 2.0 Fast) for quicker results.

See Also

• Scenes — Use generated images as video references
• Text Nodes — Generate prompts for image generation
• Draw Nodes — Sketch visual guidance for images
• Upscaler Nodes — Enhance generated images
• Models & Specifications — Complete model details
• Global Context — Project-wide context settings
• Autofill — Automatically populate prompts