# EclecticEuphoria_Flux_Unet_Models_v3

I believe it took 7 revisions to finally get a V3.  It should have better anatomy for the nsfw stuff, not tested much as I don't do much nsfw images. It still needs improving or just a good prompt, most of work went into skin texture (removing the plastic smooth look) and targeting certain blocks with reality.  The workflow image you can drag and drop into comfyui.  It's modular so you can turn things off and on, might need several new nodes, you need the dual clip and flux vae for these models.  Has not been tested outside of comfyui, need to get forge installed.

![](workflow_Flux_v3_gguf_n_mixedcast.png)
![](Flux_v3_20240906143726_USDBB_00001_.png)
![](Flux_v3_20240906_tempm_00001_.png)
![](Flux_v3_20240906141217_USDBB_00001_.png)

[Civitai Gallery of Flux Images from this model](https://civitai.com/collections/2449556)


# EclecticEuphoria_Flux_Unet_Models_v2 (Initial Release to Public)
The models are Unet models, so you need the dual clip and flux vae.  The model is a merge of models and trained with 34 Loras.  With the gguf Q8 I like to use a refiner with a SD1.5 model. Picture below, one on right is a refined version of the original

![](gguf_Q8.png)

![](20240825132439_USDB_00001_.png)

![](20240825142059_USDB_00001_.png)

![](20240825183424_USDB_00001_.png)