Introduction

Gen-4 Image is our most advanced base model for image generation, offering unprecedented stylistic control and visual fidelity.

This guide covers accessing Gen-4 Image in your dashboard, the available settings and how to use them, and how to reiterate your results to receive your desired output.

Article highlights

• Simple prompts work well, but more advanced prompts yield more control
• Enable the Aesthetic Range setting for more creatively diverse results in an image set
• Use the Vary action to create a new image set based on a single output

Related links

• Gen-4 Image Prompting Guide – Prompting best practices, examples, and more
• References Guide — Detailed overview of creating consistent characters and scenes using References

Spec Information

Step 1 – Drafting the Prompt

Begin by navigating to Generate Image from your Runway Dashboard to create a new session.

From an existing session, you can switch to image mode by selecting Image from the top of the canvas.

Either method will bring you to the text to image canvas that includes the prompting box and other settings:

Draft your text prompt in the prompting box. Simple prompts work well, but more advanced prompts will yield more stylistic control. See the Gen-4 Image Prompting Guide for examples, inspiration, and more details on prompting. 

Alternatively, select Generate Prompt to automatically describe an uploaded image. The model will analyze your image and generate a prompt describing the subject, scene, and style.

Gen-4 Image also supports References, which allow you to generate consistent character, scenes, and more by using an input image. To learn more, see the References Guide.

After drafting your prompts, you’re ready to configure the additional settings.

Step 2 – Configuring the Settings

Gen-4 Image currently supports the following settings before generating:

• Aspect ratio: Choose the dimensions of the image outputs
• Quantity: Choose between a single or batch of 4 images
• Resolution: Choose between 1080p or 720p resolution
• Style: Apply a curated style preset
• Aesthetic Range: Configure the creative variation between images in a generation set
• Settings: Configure a randomized or fixed seed (seeds are randomized by default)

Understanding the Aesthetic Range setting

The Aesthetic Range setting is powerful for exploring creative concepts in an image set without making adjustments to your prompt.

This setting can be configured to a value between 0 to 5, where 0 represents no variation from the input prompt and 5 will result in more dramatic aesthetic range.

Prompt	Aesthetic range value	Outputs
cinematic photograph of a human made entirely of white hologram static in a field of flowers at night. dynamic pose	Off
	3
	5

Step 3 – Generating the Image

After configuring the prompt and settings, you’re now ready to generate your image set. Click the purple Generate button to start the generation.

Your generations will begin processing on the right-hand side of the canvas, and will be scrollable through your session as you continue to generate.

If you receive a generation error, please ensure that your prompt complies with our acceptable content policies.

Once your generation is complete, you can hover over an image to reveal options to continue working with the output, such as Use and Vary:

• Use will load the image into the Generative Video tools, using Text/Image to Video by default
• Vary will immediately start a new image set generation in the style of the selected output

If you're ready to bring your image to life by transforming it into a video, select the Use option and learn more about our Generative Video tools on the Gen-4 Models.

Keep reading to learn how to further perfect your images through iteration.

Next steps

There are three options for iterating on your outputs once the generation completes:

• Making changes to the text prompt
• Using Vary to further explore a single result in an image set
• Iterating by reusing an input with References

Iterating with text prompt updates

Iterating with text prompt updates is helpful when you'd like to receive larger changes or to reinforce elements that were not conveyed to preference in the output of the image set.

As an example, say we used the following prompt for our first generation attempt to receive these results:

Prompt	Outputs
photograph of warped ornate classic buildings in a black abyss. stark white color palette with cool blue and green tones. avant-garde. vibrant red, blue, purple, jewel green.

However, we wish to see more exaggerated warping and and the buildings conjoined by a liquid material. Updating the prompt would be best to receive these dramatic changes:

Prompt

Outputs

photograph of warped ornate classic buildings in a black abyss. the buildings are in surreal shapes. the buildings curve and bend impossibly high into the black sky in a glitched swirl formation, defying gravity. stark white color palette with cool blue and green tones. distorted. warped, squished, and extremely deformed and exaggerated. avant-garde. vibrant red, blue, purple, jewel green. conjoined bubblegum stretch. hologram. encased in translucent liquid chrome with an iridescent sheen.

Iterating with Vary for structurally unique results from a single image

The Vary action allows you to further explore a single image by generating new results in a similar style, but with a different structure or composition.

Hover over a completed generation to reveal the Vary action. Clicking Vary will immediately start a new image set generation that’s based on the selected image. 

You can continue to use Vary on any output to explore new worlds and find inspiration.

Iterating with Gen-4 References

Gen-4 References allows you to take one or multiple images and create new images using characteristics, styles, characters, or objects from your reference images. You can extract a character from one image and place them in a different scene, transform character elements or environments, blend visual styles between images, or combine elements from multiple sources into a single new creation.

To learn more, see Creating with Gen-4 Image References.