🧔 Gym Bear Dad —— Asian Middle-aged 健身房熊爸 —— 亚洲中年### 🔑 触发词 / Trigger Word *<code>AS1BE2</code>**Design Concept: Focused on realistic bear uncles, this LoRA specializes in portraying the mature, robust charm of Asian middle-aged men. Through targeted fine-tuning, it strengthens thick bear-style physiques and lifelike details, aiming to generate powerful yet warm and approachable real bear dad images — especially highlighting the thick, strong body in gym/leisure settings.Core Features: Thick bear physique (thick chest, soft belly, broad shoulders, body hair), realistic Asian middle-aged facial features (authentic 40-65 years old look), steady aura with natural life traces, gym bear dad body (muscular yet soft belly combination, powerful and thick build).Main Strengths: Extremely realistic style, excellent thick physique rendering, stable hand and body proportions, rich natural expressions. Ideal for various occupations and body types of Asian middle-aged bear uncles, covering 40-65 age range, particularly excels in gym/leisure sports scenes. 🔑 Trigger Word *<code>AS1BE2</code>**- Compatible Base Model: Z-Image-Turbo - Strengthens Asian middle-aged male features + bear physique### 💡 Practical Tips 1. Natural Language Prompt Recommendation: Trained mainly on natural language supporting models (Z-Image-Turbo), best results come from using LLMs (ChatGPT, Claude, Gemini, Grok, Qwen) to polish prompts. Simply describe “the Asian gym bear dad scene you want” and let the LLM expand it into high-quality text. 2. Civitai Online Generation Note: Online generation works okay for Z-Image-Turbo, but strongly recommend local generation on your own PC for much better results and full LoRA strength/details. 3. Trigger Word Reminder: The trigger may slightly differ from the model name, but the latter part of the model name is usually the correct trigger. 4. To enhance Asian features → add at the end: <code>east asian features, black hair, monolids</code> 5. To make it more bear/thick → add: <code>prominent dad bod, thick waist, body hair, huge pecs</code> 6. Pose recommendations: Directly write in prompt like <code>standing confidently, sitting relaxed on sofa</code> 7. If face details are insufficient → add <code>highly detailed, 4K, ultra realistic, photorealistic, extremely detailed face</code> or lower LoRA weight to 0.75 and use ADetailer.Recommended Generation Settings: - Sampler: Euler - Steps: 9 - CFG Scale: 1 - LoRA Weight: 0.8-1.0- Resolution: 1024x10244K Version Note: Some of my models are trained with 4K datasets and support higher resolutions (recommended 1328x1328 or above). Details and skin texture will be significantly better. Pair with Hires fix.---Welcome to use and provide feedback. If helpful, feel free to support me with BUZZ. Thank you!I organize LoRAs via Collection. If you’re interested in this type of Asian middle-aged bear / Bara Daddy

Gym Bear Asian Dad : AS1BE2

zimageTurbo-AS1BE2_c1-st3000.safetensors

v1.0(Zimage turbo ZIT)


# 

## Model Introduction

This model is trained using the [Civision/model training](https://modelscope.ai/civision/modelTraining) environment and computing resources provided by ModelScope.

* Model Type: LoRA
* Base Model: [Tongyi-MAI/Z-Image-Turbo](https://modelscope.ai/models/Tongyi-MAI/Z-Image-Turbo)
* Training Code: [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* Data amount：20
* Total training steps：4000
* License: Apache-2.0

## Inference Code

Install [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio):

```bash
pip install diffsynth
```

Inference:

```python
from diffsynth.pipelines.z_image import ZImagePipeline, ModelConfig
import torch

pipe = ZImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="Tongyi-MAI/Z-Image-Turbo", origin_file_pattern="transformer/*.safetensors"),
        ModelConfig(model_id="Tongyi-MAI/Z-Image-Turbo", origin_file_pattern="text_encoder/*.safetensors"),
        ModelConfig(model_id="Tongyi-MAI/Z-Image-Turbo", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    tokenizer_config=ModelConfig(model_id="Tongyi-MAI/Z-Image-Turbo", origin_file_pattern="tokenizer/"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="jlhe2016/zimageTurbo-AS1BE2", origin_file_pattern="zimageTurbo-AS1BE2_c1-st4000.safetensors"))
prompt = "a cat"
image = pipe(prompt=prompt, num_inference_steps=8, cfg_scale=1)
image.save("image.jpg")
```


Files with SHA256: 4d5cd033469999c40e6a52406c3a89529fdef451786c7c553d1a1c31194c8471