Black Forrest Labs released recently a bunch of new tools for Flux.1 models, this include: Flux.1 Fill, Flux.1 Redux, Flux.1 Depth and Flux.1 Canny. I this post I will focus on Redux which behaves like IPAdapters if you are familiar with those and can be used to remix images and text prompts together.
Redux model is small and light which makes it easy to run locally on your GPU as it only weighs 123 Mb in size but it utilises SigClip_Vision model that interprets the image you are supplying. Before you get started in playing with these always make sure that when you try something new your ComfyUI is updated so you have the latest supported features.
So in order to run this you need to download:
- Flux1 Redux Dev model and place it under
ComfyUI/models/style_modelsfolder - Sig Clip Vision 384 and place it in your
ComfyUI/models/clip_visionfolder
You can also use this default workflow provided by download this PNG image to load into your ComfyUI
This will get you started with using Redux and experimenting with it.
I found that the text prompt does not have much affect on the output unless your dial down the strength of the image(s) using the Strength setting in the Apply Style Model node (if you don’t see a strength value you are not on the latest ComfyUI version – update).
I chained multiple image inputs in the original workflow to create a 3-image blend setup that will create the final image using the 3 inputs. I grouped them for ease of bypassing in case you only want to use two or one. I also then added LoRA node which let’s you achieve a particular style if you wish to include a LoRA. Of course you can bypass any nodes that you don’t want to use.
Interesting in trying out my workflow then you can download it here. Support my blog by buying me a coffee or checkout my shop.
Flux.1 Redux 3-Image Workflow (91 downloads )If you'd like to support our site please consider buying us a Ko-fi, grab a product or subscribe. Need a faster GPU, get access to fastest GPUs for less than $1 per hour with RunPod.io