modele/Surrealidescent - v1

Surrealidescent - v1

9/1/2025

10:37:57 AM

Powiązane słowa kluczowe i tagi

aesthetic,flux.1 d,iridescent,iridescent style,lora,markury,photorealim,photorealism,stable diffusion model,style,surreal photo,surreal photography,surrealidescent,surrealism,v1

A surreal tree made entirely of shimmering glass branches and iridescent foliage set against a bright blue sky with fluffy clouds.

A surreal iridescent octopus with sparkling tentacles floating gracefully through a galaxy of glowing stars and colorful cosmic orbs.

A detailed portrait of a female elf warrior with long flowing blonde hair, wearing ceremonial metal armor and a jeweled diadem, holding an ornate sword.

A female cyborg with long black hair and cybernetic limbs stands in a misty futuristic city illuminated by soft neon lights.

A stylish anime girl holding a katana sword, dressed in a sheer white blouse and shiny black vinyl shorts, illuminated by contrasting neon pink and blue lighting in a retro noir style.

Close-up portrait of a woman wearing an intricate crystal headpiece with sharp, angular shapes and futuristic design, featuring a dark background with sparkling reflections.

Zalecane negatywne podpowiedzi

(not used)

Zalecane parametry

samplers

euler_beta

steps

cfg

resolution

832x1216

Wskazówki

This LoRA skews generations toward photorealism with shorter prompts rather than verbose descriptions.

For enhanced skin details, consider mixing this LoRA with another model focused on that attribute.

Najważniejsze informacje o wersji

Trained with SimpleTuner on 1x RTX4090

Training steps: 4500
Learning rate: 2e-05
Effective batch size: 2
Optimizer: optimi-lion
Quantised: Yes: int8-quanto
LoRA Rank: 16
LoRA Dropout: 0.1

This is a pretty simple LoRA trained on my pinterest homepage. It's full of those renders of ice cubes and fruit, sparkly food, and cats doing weird stuff.

It's also full of surreal photography, which is hard to categorize as strictly a style when it's quite dependent on the content and context, but you know it when you see it.

It's a big ol' mix and I think of it as like that personalize flag on Midjourney. It generally skews generations toward photorealism with shorter prompts than usual (as opposed either writing verbose prompts for photorealistic detail or using a short prompt and getting that smooth synthetic look Flux is known for). With that being said, it doesn't actually add things like skin details since that wasn't the focus of the dataset, so you may want to mix it with another LoRA for that.