Model description
🧩 LoRA Name: Bella_Swan_Character
> Base Model: Tongyi-MAI / Z-Image-Turbo
> LoRA Type: Character / Film Character
> Trigger Word(s): Bella Swan (name-based prompting recommended)
> Recommended Weight: 0.6 – 0.85 (sweet spot ≈ 0.75)
✨ Overview
> This LoRA is trained to capture the visual identity and emotional tone of Bella Swan, focusing on her understated appearance, natural expressions, and quiet, introspective presence. Optimized for ZimageTurbo, it emphasizes grounded realism over exaggeration, preserving a soft, cinematic look across varied lighting and environments.
The model is well suited for moody portraits, narrative stills, and emotionally driven scenes, reflecting the character’s reserved demeanor, subtle vulnerability, and realistic, unglamorous styling.
🎨 What This LoRA Is Good At
> Consistent facial identity with natural proportions
> Subtle, reserved expressions and emotional nuance
> Soft, realistic skin rendering with minimal makeup
> Works well in low-contrast, overcast, or indoor lighting
> Strong performance in cinematic, story-focused compositions
⚙️ Usage Notes
> Best results with portrait, medium-shot, and cinematic framing
> Prompting “Bella Swan” is sufficient; avoid stacking other character LoRAs
> Performs especially well with soft daylight, overcast lighting, and muted color palettes
> Guidance scale ≤ 1 is recommended for native ZimageTurbo behavior
🧪 Example Prompt
Bella Swan, cinematic portrait, soft natural light,
introspective expression, muted colors, realistic skin texture
📐 Training Details
> Dataset Size: Captioned image set
> Training Resolutions: 512 / 768
> Total Steps: 4300
> 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
> Heavy glam makeup or extreme fashion styling may reduce character accuracy
> Designed for grounded, realistic, and narrative-driven imagery rather than fantasy-heavy scenes
Trigger words
You should use Caucasian Woman to trigger the image generation.
You should use Bella to trigger the image generation.
Download model
Download them in the Files & versions tab.
