Downloads · 30 days
39
10% of all-time downloads
JunkieMonkey69/StefanieJoosten_ZimageTurbo
StefanieJoosten_ZimageTurbo is a text-to-image model from JunkieMonkey69. Use it when you need an image from a text prompt. It is set up for diffusers. The card lists the license as apache-2.0.
Downloads · 30 days
39
10% of all-time downloads
All-time downloads
395
Public
Repo size
409 MB
Likes
4
Public
Click a slice to open those files.
.safetensors170 MB · 71%
From the Hugging Face model README
🧩 LoRA Name: Stefanie_Character
> Base Model: Tongyi-MAI / Z-Image-Turbo > LoRA Type: Character / Original Person > Trigger Word(s): Stefanie (name-based prompting recommended) > Recommended Weight: 0.6 – 0.85 (sweet spot ≈ 0.75)
✨ Overview
> This LoRA is trained to capture the facial identity, proportions, and expressive range of Stefanie, focusing on natural realism and emotionally grounded presence. Optimized for ZimageTurbo, it preserves consistent likeness across casual, cinematic, and lifestyle-oriented scenes, with particular strength in intimate portraits and urban night settings.
The model balances softness and realism, rendering natural skin texture, subtle expressions, and believable body language, making it well suited for narrative portraits, candid moments, and cinematic lifestyle imagery.
🎨 What This LoRA Is Good At
> Consistent facial identity with hazel eyes, high cheekbones, and dimples > Natural skin texture with soft, realistic lighting response > Strong performance in both indoor lifestyle and night-time urban scenes > Expressive, candid poses with relaxed body language > Works well for cinematic, casual, and emotional portrait compositions
⚙️ Usage Notes
> Best results with portrait, mid-shot, and upper-body framing > Prompting “Stefanie” is sufficient; avoid stacking other strong character LoRAs > Performs especially well with soft lighting, city bokeh, and intimate indoor setups > Guidance scale ≤ 1 is recommended for native ZimageTurbo behavior
🧪 Example Prompt Stefanie, cinematic portrait, soft ambient light, natural expression, shallow depth of field, realistic skin texture
📐 Training Details
> Dataset Size: 37 images (fully captioned) > Training Resolutions: 512 / 768 > Total Steps: 3,700 > Batch Size: 1 > Content vs Style: Balanced > Noise Scheduler: FlowMatch > Optimizer: AdamW 8-bit > Learning Rate: 1e-4 > Precision: BF16 > Text Encoder Training: Disabled > UNet Training: Enabled > Quantization: QFloat8 (model + text encoder) > Architecture: ZImage Turbo LoRA (LoKR full-rank enabled)
⚠️ Limitations / Notes
> Not intended for anime or heavily stylized cartoon outputs > Extreme facial distortion or exaggerated wide-angle lenses may reduce likeness > Heavy glam makeup or extreme fashion styling may reduce character fidelity
Download them in the Files & versions tab.