RunwayML Gen4 Image

Generate consistent characters and scenes using reference images.

Model Overview

Gen-4 Image and Gen-4 Image Turbo are advanced text-to-image models that leverage reference images to maintain character and location consistency across generations. They allow you to transform lighting, poses, settings, and styles while preserving the visual identity of your subjects.

Best At

• Maintaining character and facial consistency across multiple generations.
• Ensuring scene and environment consistency.
• Creating variations of a character or scene with different lighting, poses, or styles.
• Product mockups and character design where consistency is key.

Limitations / Not Good At

• May struggle with highly complex or abstract prompts without clear visual references.
• Extremely fine details in reference images might not always be perfectly replicated.

Ideal Use Cases

• Illustrating blog posts with a consistent character.
• Creating concept art for games or films with recurring characters and locations.
• Generating product mockups in various settings.
• Developing storyboards with consistent visual elements.

Input & Output Format

• Input: Text prompt, optional seed, aspect ratio, resolution, up to 3 reference images, and optional reference tags.
• Output: Generated image (URI format).

Performance Notes

• Gen-4 Image Turbo is optimized for speed and cost-effectiveness, being 2.5x faster than the standard Gen-4 Image model.