FLUX.1 Tools is a set of tools for Flux.1 users, and is released as a module with four functions: 1 Fill, Depth, Canny, and Redux. This function already works in Stable Diffusion, but now Flux.1 can be used in more diverse ways for high-quality images.
FLUX.1 Tools is a set of tools for Flux.1 users, and is released as a module with four functions: 1 Fill, Depth, Canny, and Redux. This function already works in Stable Diffusion, but now Flux.1 can be used in more diverse ways for high-quality images.
FLUX.1 Tools is a set of tools for Flux.1 users , and is released as a module with four functions: 1 Fill, Depth, Canny, and Redux. This function already works in Stable Diffusion, but now Flux.1 can be used in more diverse ways for high-quality images.
Of the four tools released, I'm particularly drawn to the Redux feature, which is an image reconstruction or recombinant function.
Four editing tools in Flux.1
First, let me briefly explain each function provided in a modular manner.
Fill : Modifies part of the image (In-painting) or creates by extending outside the image (Out-painting)
Depth : Determine the depth of the image, layer the background and foreground, and modify the image based on that depth.
Canny : Finds all the boundary lines in the image and creates an image based on that structure.
Redux : Reassemble and restyle elements in different ways using image elements and text prompts.
In fact, these editing features, which started in Stable Diffusion, have been added to Flux.1, Midjourney, It is also implemented in various tools such as Dreamina and Ideogram .
In Midjourney, it corresponds to the Editor, Pan, Zoom, and Retexture functions.
In the free tool Dreamina, editing functions can be said to correspond to Retouch, Inpaint, Expand, and Image Reference (Face/Style/Edge/Depth). ( Refer to AI Salson 2024.11.21)
Stable Diffusion and Flux , which are always open source and can quickly test new and diverse features.This is a series. This time, it shows a really effective function.
However, the fact that users must download models or modules for each update, install them locally, and test them with their own know-how is still an inconvenient obstacle to rapid popularization.