Sign In

Rebel_Essence_ZiT

7

Download

1 variant available

SafeTensor

60.88 MB

Verified:

Type

LoRA

Stats

96

7

Reviews

Published

May 23, 2026

Base Model

ZImageTurbo

Hash

AutoV2
EDA4BE52CC

Trigger Words

rebelessence

Like's Button is free

RebelEssence_ZIT — Punk/Gothic Editorial Style LoRA (ZImage Turbo)

A modular, stable, clean and highly coherent Style LoRA.

RebelEssence_ZIT is a Style LoRA designed to bring a dark fashion, punk/gothic, editorial aesthetic into your generations. It delivers clean skin, precise makeup, metallic accessories, and a controlled alternative mood without overpowering the base model.

The LoRA was trained using a modular 10‑folder dataset structure, with each folder dedicated to a specific visual component:

  • Reference (face, skin, proportions)

  • Hair (texture, volume, consistency)

  • Makeup (gothic/punk, eyeliner, smokey eyes)

  • Outfit (leather, mesh, dark fashion)

  • Accessories (chokers, chains, metal details)

  • Poses (natural and dynamic posture)

  • Full Body (proportions and silhouette)

  • Lighting (neon, soft contrast, dramatic light)

  • Backgrounds (studio, urban dark, industrial)

  • Close-ups (skin texture, eyes, fine details)

This structure gives the model an exceptionally balanced bias: no drift, no forced identity, no unwanted facial imprinting, and no contamination. It responds to prompts in a clean, controlled, and predictable way.

LoRA Strength (Weight)

RebelEssence_ZIT remains stable even at high strengths.

  • Recommended range: 0.7 → 1.0

  • 0.7 = softer integration, more natural look

  • 0.8–0.9 = sweet spot, strong and coherent style

  • 1.0 = full identity with no artifacts

The LoRA holds 1.0 without collapsing, without glitches, and without plastic skin — a rare behavior for ZImage Turbo.

Training Details

  • Base: ZImage Turbo

  • Rank: 12

  • Alpha: 1

  • Learning Rate: 0.000085

  • Steps: 1600

  • Batch Size: 2

  • Weight Decay: 0.0001

  • Dataset: Modular 10‑folder structure

  • Trigger Word: rebelessence

The training was calibrated to achieve:

  • strong but non-invasive identity

  • consistent style without rigidity

  • zero ethnic drift

  • zero artifacts

  • zero contamination

  • zero plastic skin

What You Can Create with RebelEssence_ZIT

  • dark fashion editorial portraits

  • punk/gothic beauty shots

  • full-body leather/mesh outfits

  • neon lighting (blue/purple)

  • urban dark scenes

  • high‑definition close-ups

  • alternative/punk metal aesthetics

  • ethnic variations (asian, african, latina, caucasian) without drift

The model respects ethnic prompts and does not override them — a sign of a clean dataset and controlled bias.

Why It’s a Modular Style LoRA

Thanks to the 10‑folder structure, RebelEssence_ZIT:

  • does not impose a fixed face

  • does not force a specific outfit

  • does not repeat poses

  • does not lock backgrounds

  • does not collapse into one lighting style

It interprets your prompt instead of copying the dataset. This makes it extremely versatile and production‑ready.