This is a very straight-forward and simple workflow that I use often for NSFW generations with the Flux 2 Klein 9b (Fp8 scaled version).

• The workflow employs two separate image inputs for character replacement/editing.

Simple prompts work like the following:

- "The female from image1 replaces the female inside image2"

- "The female from image1 is wearing the outfit from image2"

• You can also disconnect the 2nd image input from the Positive Prompt Text Encoder, to only work with a singular image input.

My generations are often run at 10 steps, use a combination of Euler - Simple, Denoise between .75 -1 depending on how far I'm straying from my original reference (high value for bigger changes), and CFG at 1 (negative prompt disabled because of this).

Here is a list of models that I use with this workflow:

Diffusion Model:

• https://huggingface.co/black-forest-labs/FLUX.2-klein-9b-fp8

Load Clip/Text Encoder:

• https://huggingface.co/Comfy-Org/vae-text-encorder-for-flux-klein-9b/blob/main/split_files/text_encoders/qwen_3_8b_fp8mixed.safetensors

Vae:

• https://huggingface.co/black-forest-labs/FLUX.2-small-decoder/blob/main/full_encoder_small_decoder.safetensors

Here are the 2 LoRAs that I use for most of my NSFW generation:

• https://civarchive.com/models/2416142/snofs-sex-nudes-and-other-fun-stuff-flux-2-klein-9b-base-and-distilled?modelVersionId=2985440

• https://civarchive.com/models/1662740/lenovo-ultrareal?modelVersionId=2682771