Chinese-Female-NudeShot-NSFW-LoRA-v2

Chinese Female NudeShot NSFW LoRA v2

Chinese Female NudeShot NSFW LoRA v2 is a large-scale photorealistic LoRA for Qwen/Qwen-Image-2512 Text-to-Image. It was trained on 108,974 images to strengthen realistic adult Chinese female photography, natural skin rendering, body proportions, varied poses, lighting, environments, and photographic composition.

Compared with v1, this release uses a much larger and more varied training set. The dataset was prepared without a single fixed visual bias, allowing the LoRA to respond more naturally to prompt-defined subjects, clothing, poses, locations, lighting, and camera framing.

Adult-content notice: This repository contains explicit adult material and is intended only for adults. All people described in the examples are adults.

---

✨ Features

• Photorealistic adult Chinese female imagery
• Large and varied 108,974-image training set
• Natural skin texture, lighting, shadows, and depth
• Broad support for indoor, outdoor, portrait, and full-body compositions
• Reduced dependence on one fixed face, pose, background, or photographic style
• Designed and tested for Qwen-Image-2512 T2I

---

🖥️ Training Details

Hardware

• GPU: NVIDIA RTX PRO 6000 Blackwell 96GB

Dataset and Network

• Training images: 108,974
• Linear rank: 32
• Linear alpha: 32
• Batch size: 8
• Training steps: 25,000

---

⚠️ Base Model Compatibility

This LoRA is intended for:

• Qwen/Qwen-Image-2512 for Text-to-Image generation

Compatibility with other Qwen-Image checkpoints or architectures is not guaranteed. An incompatible base model may produce weak effects, artifacts, incorrect colors, or a loading error.

---

🔍 Scope and Limitations

This LoRA learns visual patterns from a large photographic dataset. Prompt wording, seed, resolution, LoRA weight, inference steps, and the selected base model all affect the result.

Although the larger dataset improves variety and generalization, the model does not guarantee perfect human anatomy. Difficult poses, occlusion, unusual camera angles, and tightly framed images may still produce incorrect hands, limbs, proportions, or intimate anatomy. Review generated results carefully and regenerate or edit visible defects when necessary.

---

🖼️ Examples

The following six images were generated with this LoRA. All people described in the prompts are adults.

---

Example 1

Prompt

A woman with dark hair is lying on a green pool table, with her head turned towards the camera. She is wearing a white shirt that is open, revealing her bare chest. She is holding a red pool ball in her right hand. There are three black lamps hanging above her, casting a warm light on her body. The background is dark, with a black wall and a framed picture on it. The pool table has a green felt surface and a wooden edge. There is a pool cue lying on the table to the left of the woman.

---

Example 2

Prompt

This is a photograph of a young adult Asian woman with long, straight black hair and bangs. She is nude and lying on her side on a bed with white sheets. Her right leg is bent at the knee and lifted, while her left leg is extended. Her left hand is resting on her hip, and her right hand is placed on her thigh. She has a natural, unshaven pubic area. The background features a patterned wallpaper with black silhouettes of people dancing and various circular designs in shades of blue and gray. To the right, there is a dark purple curtain. The overall setting appears to be a bedroom with a modern, slightly eclectic decor.

---

Example 3

Prompt

The image features a young adult Asian woman standing in front of a window with a cityscape view in the background. She has long, straight brown hair tied back with a red hair tie, and her face is turned slightly towards the camera, showcasing her profile. The woman is wearing a black sleeveless top that appears to be made of a lightweight, possibly sheer material. Her right hand is raised above her head, with fingers gently touching her hair, while her left hand rests on her hip. The lighting is soft, creating a gentle contrast between the bright window and the darker interior. The overall atmosphere of the image is serene and contemplative.

---

Example 4

Prompt

A nude adult woman stands in shallow water at the edge of a beach. She has long, dark hair and is facing slightly to the left of the camera. Her arms are raised, and her hands are placed on her head, possibly adjusting her hair. She is wearing a white bracelet on her right wrist. The water is clear and turquoise, with small waves lapping at the shore. In the background, the ocean stretches out to the horizon, and a small boat is in the distance on the left side of the image. The sky is a clear, bright blue. The beach is sandy and appears to be relatively empty.

---

Example 5

Prompt

An adult person with fair skin is seated on a light-colored, possibly marble, countertop. They are wearing a pink lace lingerie set, which includes a bra and panties. The panties are low-rise and feature a garter-belt design with straps that extend along the thighs. The person has long, straight brown hair that cascades down their back. Their legs are spread apart, and they are wearing red nail polish on their toenails. The background includes a white sink with a silver faucet.

---

Example 6

Prompt

This is a photograph of a young adult Asian woman sitting nude in a small, weathered wooden rowboat on a calm, greenish body of water. The woman has long, dark hair and is wearing silver sandals. She is positioned on the right side of the boat, leaning back slightly with her legs crossed, and looking towards the camera with a neutral expression. Her body is turned slightly to the left, and her arms are resting on her knees. The boat is made of dark, aged wood with visible wear. On the left side of the boat, an open blue parasol with pink and white floral designs rests on the edge and is tied to the boat with a rope. The still water reflects the boat and the woman, while a blurred green shoreline suggests a natural, secluded setting. Soft lighting suggests early morning or late afternoon.

---

🚀 Reference Code — Qwen-Image-2512 T2I

This example loads Qwen-Image-2512 in BF16 and uses the official 4-step Lightning LoRA. A high-memory CUDA GPU is recommended.

Installation

pip install -U diffusers transformers accelerate peft safetensors

Python

import math

import torch
from diffusers import DiffusionPipeline, FlowMatchEulerDiscreteScheduler


BASE_MODEL_ID = "Qwen/Qwen-Image-2512"
LIGHTNING_MODEL_ID = "lightx2v/Qwen-Image-2512-Lightning"
LIGHTNING_WEIGHT_NAME = (
    "Qwen-Image-2512-Lightning-4steps-V1.0-bf16.safetensors"
)

LORA_MODEL_ID = "perpetual3x/Chinese-Female-NudeShot-NSFW-LoRA-v2"
LORA_WEIGHT_NAME = "chinanudeshotv3.safetensors"

DEVICE = "cuda"
DTYPE = torch.bfloat16


scheduler_config = {
    "base_image_seq_len": 256,
    "base_shift": math.log(3),
    "invert_sigmas": False,
    "max_image_seq_len": 8192,
    "max_shift": math.log(3),
    "num_train_timesteps": 1000,
    "shift": 1.0,
    "shift_terminal": None,
    "stochastic_sampling": False,
    "time_shift_type": "exponential",
    "use_beta_sigmas": False,
    "use_dynamic_shifting": True,
    "use_exponential_sigmas": False,
    "use_karras_sigmas": False,
}

scheduler = FlowMatchEulerDiscreteScheduler.from_config(scheduler_config)

pipe = DiffusionPipeline.from_pretrained(
    BASE_MODEL_ID,
    scheduler=scheduler,
    torch_dtype=DTYPE,
).to(DEVICE)

pipe.load_lora_weights(
    LIGHTNING_MODEL_ID,
    weight_name=LIGHTNING_WEIGHT_NAME,
    adapter_name="lightning",
)
pipe.fuse_lora(adapter_names=["lightning"], lora_scale=1.0)
pipe.unload_lora_weights()

pipe.load_lora_weights(
    LORA_MODEL_ID,
    weight_name=LORA_WEIGHT_NAME,
    adapter_name="chinanude_v2",
)
pipe.set_adapters(["chinanude_v2"], adapter_weights=[1.0])

prompt = (
    "an adult Chinese woman, realistic photography, "
    "natural skin texture, soft window light, detailed environment"
)

negative_prompt = (
    "low quality, blurry, distorted anatomy, malformed hands, extra limbs, "
    "duplicated body parts, plastic skin, CGI, watermark, text"
)

generator = torch.Generator(device=DEVICE).manual_seed(42)

image = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=928,
    height=1328,
    num_inference_steps=4,
    true_cfg_scale=1.0,
    generator=generator,
).images[0]

image.save("chinese_female_nudeshot_v2.png")

---

🧩 Low-VRAM Inference (16–24 GB)

Most VRAM is consumed by the Qwen-Image-2512 base model and its text encoder, not by this LoRA. Users with lower-memory NVIDIA GPUs can combine 4-bit quantization, CPU offloading, VAE tiling, batch size 1, and a lower output resolution.

Low-VRAM Installation

pip install -U diffusers transformers accelerate peft safetensors bitsandbytes

Low-VRAM Python

import torch
from diffusers import DiffusionPipeline
from diffusers.quantizers import PipelineQuantizationConfig


BASE_MODEL_ID = "Qwen/Qwen-Image-2512"
LORA_MODEL_ID = "perpetual3x/Chinese-Female-NudeShot-NSFW-LoRA-v2"
LORA_WEIGHT_NAME = "chinanudeshotv3.safetensors"

quant_config = PipelineQuantizationConfig(
    quant_backend="bitsandbytes_4bit",
    quant_kwargs={
        "load_in_4bit": True,
        "bnb_4bit_quant_type": "nf4",
        "bnb_4bit_compute_dtype": torch.bfloat16,
    },
    components_to_quantize=["transformer", "text_encoder"],
)

pipe = DiffusionPipeline.from_pretrained(
    BASE_MODEL_ID,
    torch_dtype=torch.bfloat16,
    quantization_config=quant_config,
)

pipe.load_lora_weights(
    LORA_MODEL_ID,
    weight_name=LORA_WEIGHT_NAME,
    adapter_name="chinanude_v2",
)
pipe.set_adapters(["chinanude_v2"], adapter_weights=[1.0])

pipe.enable_model_cpu_offload()
pipe.vae.enable_tiling()

prompt = (
    "an adult Chinese woman, realistic photography, "
    "natural skin texture, soft photographic lighting"
)

negative_prompt = (
    "low quality, blurry, distorted anatomy, malformed hands, extra limbs, "
    "duplicated body parts, watermark, text"
)

generator = torch.Generator(device="cuda").manual_seed(42)

image = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=768,
    height=1024,
    num_images_per_prompt=1,
    num_inference_steps=20,
    true_cfg_scale=3.0,
    max_sequence_length=256,
    generator=generator,
).images[0]

image.save("chinese_female_nudeshot_v2_low_vram.png")

If this still causes an out-of-memory error, replace pipe.enable_model_cpu_offload() with pipe.enable_sequential_cpu_offload(). Sequential CPU offloading uses less GPU memory but is substantially slower.

---

🔞 Responsible Use

This model is intended exclusively for lawful, consensual, adult creative work. Do not use it to:

• create or sexualize minors or youthful-looking persons
• create non-consensual or exploitative sexual content
• depict an identifiable real person without permission
• harass, threaten, exploit, or defame another person
• create illegal content or bypass platform rules

Users are responsible for ensuring that their prompts, source material, and generated content comply with applicable laws and platform policies.

---

License & Copyright Notice

This LoRA may be used for personal and non-commercial purposes only. See the repository LICENSE file for the complete terms.

For commercial licensing inquiries, please contact the author.