Z-Image Turbo (ZIT) — Optimized Multi-Phase Workflow

EchoSingularityEchoSingularity
35
855
high-resolutioncinematiccomfyuiworkflowtoolupscalerlora loader12gbseedvr2face-detailerz imagezitz-image-turbocreator-toolsclean-workflow12gb-vrammaterial-realismres2detail-daemon
v1.2v1.1v1.0
model-image-0
model-image-1
model-image-2
model-image-3
model-image-4
model-image-5
model-image-6
model-image-7
model-image-8
model-image-9
model-image-10
model-image-11
model-image-12
model-image-13
model-image-14
model-image-15
model-image-16
model-image-17
model-image-18

Details

Download Files (1)

Model description

Overview

This is a multi-phase Z-Image Turbo workflow designed to extract maximum detail, edge fidelity, and material realism on 12GB VRAM GPUs. This workflow also includes seed variance to the conditioning so that outputs with the same prompt have more variety similarly to SDXL, Pony, IL models. (Each photo pair in the showcase gallery is the exact same prompt) This workflow uses controlled sigma shaping, Res-2 samplers, and phased refinement passes to stabilize detail while avoiding common ZIT artifacts like:

  • Over-etched hair

  • Shimmering edges

  • Checkerboard blockiness

  • CFG-induced harshness

The result is clean, high-contrast outputs that scale well across portraits, fashion, cinematic scenes, and hard-surface material tests.

Key Features

Detail Daemon Sigma Control

  • Custom sigma curve shaping via Detail Daemon Graph + Sampler

  • Prevents early over-sharpening

  • Preserves skin gradients while locking edges late in the pass

Low-CFG Stability (3.0 Primary Pass)

  • Primary generation operates at CFG ≈ 1.0 - 3.0

  • Keeps ZIT from “fighting” the prompt

  • Enables natural lighting and material response

Multi-Phase Refinement

  • Phase 1: Base structure & composition

  • Phase 2: Detail reinforcement (carefully bounded)

  • Phase 3: Edge cleanup + block smoothing

  • Phase 4: Final polish, contrast, and optional desaturation

Includes side-by-side phase comparison nodes for tuning and learning.

Res-2 Sampler Focus

  • Res-2 (beta) chosen over DPM++ for:

    • Cleaner hair behavior

    • Reduced sparkle artifacts

    • Better late-stage convergence

  • DPM++ intentionally avoided due to aggressive micro-contrast amplification

SeedVariance

  • Optional Seed Variance Node for:

    • Diverse outputs using the same prompt

    • Sliders control variance percent & strength

VRAM-Aware Cleanup

  • CacheCleaner + Easy-Use GPU cleanup nodes included

  • Designed for long batch sessions

  • Prevents silent VRAM creep during multi-phase runs

Hardware Target

  • 12GB VRAM GPUs (RTX 3060 / 4070 / 5070 class)

  • Not designed for 8GB cards without modification or disabling of phases

  • Scales upward cleanly on 16GB+ systems

Recommended Usage

  • Portraits & fashion

  • Cinematic lighting setups

  • Material studies (glass, metal, diamonds, liquids)

  • Showcase renders where edge quality matters

  • ZIT prompt experimentation without CFG chaos

Images made by this model