If you like the model, I will also make the FP8 version

VelvetForge_ZIT is a hybrid model developed inside the Z‑Image Turbo pipeline, created through direct LoRA integration within a flow‑based architecture using QWEN 3.4B CLIP and BF16 precision. It is not an SDXL model, does not rely on diffusion sampling, and does not support traditional schedulers. VelvetForge_ZIT operates entirely through the AuraFlow system, using low‑step sampling with deterministic behavior.

This model blends the structural consistency of Z‑Image Turbo with a velvet‑soft semi‑real aesthetic, characterized by:

• smooth transitions

• clean, non‑plastic skin

• soft, controlled lighting

• natural editorial look

• absence of diffusion‑style micro‑noise

• stable color response even at low CFG

The fusion was calibrated to preserve the physics of the flow model, avoiding drift, artifacts, and instability commonly seen in static merges. VelvetForge_ZIT uses dynamic LoRA injection during inference, with adaptive strength handling and internal normalization, ensuring consistent results even with minimal prompting.

⭐ Pipeline & Architecture

• Base Model: Z‑Image Turbo (flow‑based)

• CLIP: QWEN 3.4B

• Precision: BF16

• Sampling: AuraFlow (res_multisteps / simple)

• Optimal Steps: 8–9

• Optimal CFG: 0–1

• Compatibility: Z‑Image Turbo pipeline only

• Not compatible with: Euler, DPM++, Heun, UniPC, Karras, DDIM, or any SDXL‑based scheduler

⭐ Model Identity

VelvetForge_ZIT produces a semi‑real aesthetic — a balanced midpoint between realism and soft editorial rendering. It is not photographic and not illustrative: it is a hybrid look with smooth light, natural textures, and clean detail.

Key characteristics:

• velvet‑smooth skin

• soft cinematic lighting

• consistent results on male and female subjects

• excellent handling of hair, clothing, and materials

• clean, non‑noisy backgrounds

• high coherence even with short prompts

⭐ Recommended Usage

• Sampler: res_multisteps (recommended), simple

• Steps: 8–9

• CFG: 0–1

• Prompting: short, clean prompts without style stacking

• Negative prompt: minimal or none

• External LoRAs: not recommended (breaks flow‑based consistency)

⭐ Intended Use

VelvetForge_ZIT is ideal for:

• semi‑real portraits

• half‑body editorial shots

• fashion soft‑light compositions

• clean, natural, velvet‑toned aesthetics

• professional Z‑Image Turbo workflows