How to Create with Runway Gen-4 Image References

Maintaining visual consistency across AI-generated content has long been one of the medium’s greatest hurdles. Whether trying to keep a specific character’s face identical across different lighting conditions or keeping a commercial product design intact within a new environment, creators frequently battle random variations.

Runway’s Gen-4 Image model addresses this directly through its References workflow. Available via [site.help.runwayml.com/hc/en-us/articles/40042718905875-Creating-with-Gen-4-Image-References](https://site.help.runwayml.com/hc/en-us/articles/40042718905875-Creating-with-Gen-4-Image-References), the Gen-4 Image References tool allows creators to pull elements from one or multiple input images—such as characters, objects, styles, or layouts—and weave them into completely new generations without requiring manual fine-tuning or model training.

How the Gen-4 References Workflow Operates

At its core, the system shifts control from pure text prompting to a hybrid approach combining visual inputs and descriptive language.

To use references within Runway’s Tool mode:

1. Navigate to the Image tab under the generation panel and select Gen-4 Image (or Gen-4 Image Turbo).
2. Toggle on the References feature.
3. Upload or select up to three active reference images per generation.
4. Tag and name your references so you can easily reference them in your text prompt using the @ symbol for autocomplete.

1. Single Reference Control: Infinite Character and Object Consistency

When dealing with narrative storyboarding or recurring digital actors, using a single reference image cuts through the unpredictability of traditional text-to-image models. For optimal output, Runway recommends starting with a clear, moderately lit photograph of your subject with a neutral expression.

By pairing this reference with a descriptive text prompt, you can seamlessly transplant a character into entirely new environments, artistic treatments, or dynamic lighting scenarios while maintaining their core identity.

2. Multi-Reference and Sketch Integration: Precision for E-Commerce and Layouts

When text descriptions aren’t enough to dictate precise composition, Gen-4 supports multi-reference prompting. Creators can combine up to three sources—such as pairing a real product photograph with a rough sketch or layout map.

• The Product Source: Supplies the object data, textures, and exact subject characteristics.
• The Layout/Sketch Source: Tells the model where elements should sit on the canvas, mapping boundaries for props, blank negative space for future text, and focal positioning.

This method bypasses the limitations of lengthy prompts, which often struggle with spatial arrangement, allowing e-commerce creators and designers to enforce structured framing.

Pro Tips for Maximizing Gen-4 Image References

• Embrace a “Blank Canvas” Input: For characters and general objects, choose reference photos with even, natural lighting and neutral backgrounds. This prevents the model from baking unwanted background hues or harsh shadows into your new scene.
• Write Positive, Descriptive Prompts: Gen-4 does not support negative prompting (