The Vividator

= = = 30 Nov 2025 = = =

The Vividator for SDXL - providing much the same effect as the Flux version (see below).

Trigger is: cs_vvd

Trained on base SDXL. Works with most SDXL checkpoints.

Showcase images made with Crystal Clear XL

/model/122822?modelVersionId=133832

because it's neutral and generally nicer than base SDXL.

Steps: 30

Sampler: DPM++ 2M

Schedule type: Karras

CFG scale: 7

Size: 896x1152

Negative prompt: nude,naked,signature

= = = 29 Nov 2025 = = =

The Vividator - it makes things ridiculously, stupidly... and sometimes artfully vivid.

Generally it will boost colour saturation and turn humdrum scenes into stylish scenes - as if by someone with no sense of restraint, practicality or good taste.

It will work with short and long prompts, although with short prompts it sometimes picks an element to isolate.

It's all about presentation and lack of subtlety.

That said, I weighted it to do more to people (or people-adjacent beings) than to creatures or objects.

Shout-out and thanks to @TomTomorrow for helping me test this LoRA!

At strength 1.0, it is LOUD.

At strength 0.6 or 0.7, it is almost controllable.

At strength 0.5 or less, it can work as part of a LoRA recipe.

The Vividator for Flux is trained on Flux1.dev.

The showcase images are all with/without pictures generated on a quant of Flux1.dev in Forge:

Flux1-Dev BNB NF4 v2

/model/638187?modelVersionId=721627

The Showcse prompts are single words or short phrases. I'll add a separate post with longer prompts.

For each pic, the left half is with The Vividator, the right half uses no LoRAs.

Steps: 20

Sampler: Euler

Schedule type: Simple

CFG scale: 1

Distilled CFG Scale: 3.5

Size: 896x1152

For Forge users:

When I use the "Flux1-Dev BNB NF4 v2" checkpoint I set:

Diffusion in Low Bits = bnb-nf4 (fp16 LoRA)

If I'm using something different, such as "Flux.Krea Q8 from FP16"

/model/1731900/flux-gguf-models-fp16-conversions?modelVersionId=2073588

then I set:

Diffusion in Low Bits = Automatic (fp16 LoRA)