FLUX Modular WF

UPDATE - Feb. 8th, 2025

Uploaded the fixed version of the workflow. There was a problem with the Save Image nodes due to an update in other nodes. This version fixes it.

FLUX Modular WorkFLow is a ComfyUI workflow that works like a "Swiss army knife" and is based on FLUX Dev.1 model by Black Forest Labs.

You can use the original model files or the GGUF versions.

There are two different basic workflows for generating an image: the standard FLUX workflow (with the optional Detailer Daemon nodes) and the Super-FLUX workflow (inspired by Olivio Sarikas' workflow). The Super-FLUX wf just split the total steps into three distinct sampler generations (1/3 of the total steps on each sampler) bringing more details to the image (sometimes even too much!).

The FLUX tools are all available: Redux (IP-Adapter), Inpaint/Outpaint, and Depth/Canny (ControlNets).

Prompt management is extremely powerful: you can use JoyCaption 2 for captioning an image and generate a detailed prompt; there is a local LLM (running on ComfyUI itself!) for LLM prompting generation; you can keep up to 6 different prompts for easy and quick use; last but not least you can create an image-database of your most used prompts: save the images you like in a specific folder, and use them as prompt-database. Just upload the image you want and the prompt you used for that image (and all the metadata saved with it) will be available for your next job.

The Modules available are the following:

1) Latent Noise Injection - improve the details of the image;

2) Expression Editor - to modify the expression of your subject in portrait images;

3) ADetailer - to improve quality and detail of hands, eyes, and faces;

4) Ultimate SD Upscaler - to upscale your images with the upscale models you like;

5) Postprocess - the final retouch to your images: saturation, contrast, sharpen, grain, and apply LUT.

The workflow needs ComfyUI with Python 3.11, as the latest version of ComfyUI is based on Python 3.12 and some custom nodes are not working with 3.12. On my YouTube channel, you can find a workaround if you have windows_portable ComfyUI with Python 3.12 and want to revert back to 3.11.

If you have trouble (Out of memory errors) with the original FLUX Dev model, I suggest you to try the GGUF version of the model: Q8 is almost the same as the original one, Q6.K or Q4.K are good alternatives.

Use the following keys to navigate the Workflow: "1" for the frontend; "2" for the explanation notes and links to models to download; "3" for the backend (usually you will need this just for the first use).

How to use the workflow

The first time you upload the workflow you will need to install 35 custom nodes (unless you use some of them already for other workflows or you already used my old workflows like the v.4.3).

I suggest installing one custom node at a time, restarting ComfyUI to make sure the installation was successful, and then installing the next custom node.

Once you have the workflow ready with all the custom nodes, you will need the models files (unet, vae, gguf, clip, clipvision, stylemodel, upscale model, sam model, bbox detector, LUT...). In the workflow's notes (press "2") you will find all the links if you need to download them or if the ComfyUI Manager does not have them available for installation (always check ComfyUI Manager first!).

Before you run the workflow, you need to do one last thing: upload the images for Input Images (Primary and Secondary) and the Prompts & parameters DataBase group. You can use any image you want to run the workflow, it's just because those nodes need an uploaded image to work. Otherwise, the workflow may get stuck.

The Worflow video-guide on my YouTube channel:

1. First time use and how the Prompt modules work

2. Coming soon...

For those of you who are happy with my workflow, please consider subscribing to my Patreon (subscribers will get priority support for the wf, and also some early updates about it) or maybe you could just Buy Me a Coffee.

One last thing, if you want to share images you create with my workflow, I would really appreciate it if you post them in the Gallery below, just click on "+ Add Post" button.

Thanks

Tenofas