Flux.1 Dev Workflow with ControlNet, Face Detailer, FaceSwap, Upscale, Multimodel (Checkpoints)

This is a Flux.1 workflow. I noticed that most workflows only use one type of checkpoint model. I thought it was silly that I needed multiple workflows for the different model types.

In this workflow, all model types (AIO checkpoints, diffusion models and GGUF models) are supported through the use of switches. See the notes on what switches do what and how to use them.

ControlNet is also supported for I2I generation, Face Detailer and Ultralytics are both implemented (bypassed by default) and ReActor Fast Face Swap is included along with nodes to save a face model for later use.

Finally, an Upscaler is included. I have found that a low CFG (3.5 in the "FluxGuidance prompt" and 1.0 in the KSampler) seems to work well together with the 4x-UltraSharp model.

Settings:

Aspect Ratio: set the "aspect_ratio" widget as desired. It will override the "width" and "height" widgets automatically. if you want to use the width and height, set the "aspect_ratio" widget to custom

FluxGuidance (in the Prompt Group): 3.5

KSampler: Steps: 20, CFG: 1.0, Sampler_name and Scheduler: dpmpp_2m & sgm_uniform, seed is randomized and denoise is 1.

ControlNet: Enabled - set Apply ControlNet widgets to anything other than zero, Disable - set Apply ControlNet widgets to zero

Models Req'd:

Checkpoint: flux1-dev-fp8.safetensors

Diffusion Model: flux1-dev.safetensors

GGUF: flux1-dev-Q8_0.gguf

Clip Models: clip_l.safetensors, t5xxl_fp16.safetensors

VAE: flux_vae.dafetensors

ControlNet: flux_dev_controlnet_union.safetensors (preprocessors should automatically download when you use them)

Florence2: Florence-2-base

Upscale: 4x-UltraSharp.pth

Ultralytics/SAM: bboxEyes.pt, sam_vit_b_01ec64.pth, person_yolo8m-seg.pt

If there is anything else you'd like to see in a AIO workflow, let me know and I'll see what I can do!