Image to Image

Summary

Image to Image is an essential tool for creative evolution. It empowers transforming existing visuals, refining details, experimenting with styles, and enhancing narratives with precision. Whether you’re iterating on a concept sketch, reimagining a scene with new aesthetics, or blending multiple ideas into a cohesive composition, Image to Image unlocks new dimensions of creativity, enabling continuous exploration and storytelling through visual transformation.

Image to Image Models

Models	Description	Best For
Flux Canny [Default]	Utilizes edge detection to preserve structural composition while generating new images based on a text prompt.	Precise composition control through edge detection. Images with a lot of details.
Flux Depth	Employs depth mapping to maintain spatial relationships in the generated images for accurate perspectives.	High accuracy in perspective and object positioning through 3D spatial consistency.
Flux Redux	Produces slight variations of an input image without the need for a text prompt, allowing for easy refinement.	High creative freedom with loose compositional guidance.

Parameters

Parameter	Type	Effect on Output
Prompt	Text	The text prompt is processed through various visual information extraction models (Canny, Depth, Redux), each uniquely interpreting different features of the input image to generate new visuals.
Style	Different kinds of stylistic presets (3D, Anime, Cyberpunk, Extreme Detailer, Game Asset, Logo, Midjourney, Motion Blur, Panorama, Pixar, Pixel, Retrofuturism, Studio Ghibli, Tarot Card, Thermal Image, Vector, Victorian Drawing, Watercolor).	The style presets modify the rendering phase by influencing color schemes, textures, and artistic techniques, transforming the structural output from ControlNet into diverse visual interpretations.
Strength	0%-100%	The strength parameter alters the amount that the text influences the output image versus the source image. The higher the strength is, the more influence the text has.
Image Size	Auto, Landscape - 16:9 (576x1024), Portrait - 4:3 (768x1024), 1:1, Landscape - 4:3 (1024x768), Portrait - 9:16 (1024x576).	This changes the output size of the node with a couple of options.
Seed	Seed	The seed is a deterministic number that indexes generations from the model. It's typically randomized, but you can set a seed if there's a particular output you're looking for! Keep in mind that all parameters must be the same in order for a given seed's output to persist.

Prompt

In the Flux Model, the text prompt directs style and content across all three models.

• "Vibrant oil painting colors"

• "Pastel watercolor tones"

• "Surreal cosmic color shifts"

• "Retro VHS color grading"

• "High-contrast monochrome sketch"

• "Dreamy ethereal palette"

• "Moody film noir lighting"