WAN21 - PAWG LoRA (Model Description)
The WAN21 - PAWG LoRA is a specialized Low-Rank Adaptation (LoRA) for Stable Diffusion, engineered to enhance or generate images with a striking focus on curvaceous, exaggerated lower-body proportions—perfectly capturing the "PAWG" (Phat Ass White Girl) aesthetic. Designed for the WAN 2.1 framework, this LoRA integrates seamlessly with ComfyUI’s repackaged WAN 2.1 models, delivering photorealistic or stylized renders with voluptuous hips, toned thighs, and plump buttocks. Ideal for artists crafting intimate portraits, sultry character designs, or provocative scenes, it offers versatility for both SFW and NSFW applications (per usage guidelines).
Pair this LoRA with detailed prompts and the WAN 2.1 model suite to create tantalizing, siren-like figures bathed in dreamy or bold lighting. From serene nudes to seductive strip club vibes, WAN21 - PAWG LoRA brings luscious curves and captivating detail to life.
Key Features:Amplifies lower-body proportions (e.g., hips, thighs, and buttocks) with natural or exaggerated styling.
Optimized for WAN 2.1-based workflows in ComfyUI.
Adjustable LoRA weight (0.6–1.0) for subtle or dramatic effects.
Thrives on rich prompts specifying pose, style, and ambiance.
To harness WAN21 - PAWG LoRA in ComfyUI, download and install these WAN 2.1 repackaged model files from Hugging Face, then move them to the appropriate directories:
Text Encoder: umt5_xxl_fp8_e4m3fn_scaled.safetensors
Move: mv umt5_xxl_fp8_e4m3fn_scaled.safetensors /workspace/ComfyUI/models/text_encoders/
VAE: wan_2.1_vae.safetensors
Download: wget https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors
Move: mv wan_2.1_vae.safetensors /workspace/ComfyUI/models/vae/
Diffusion Model: wan2.1_t2v_14B_bf16.safetensors
Download: wget https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_14B_bf16.safetensors
Move: mv wan2.1_t2v_14B_bf16.safetensors /workspace/ComfyUI/models/diffusion_models/
These files provide the foundation for text encoding, variational autoencoding, and diffusion, ensuring compatibility and high-quality outputs with the WAN21 - PAWG LoRA.
Example Prompts:"A photo-realistic shoot from below, featuring a nude woman with long blonde hair standing in the middle of the frame, facing away from the viewer, with her arms raised above her head. Her body is slender and toned, with a focus on her back and buttocks. She appears to be a young adult, with fair skin and a slender physique. Her hair is styled in a long, straight style, and her eyes are closed, giving her a serene expression. Her arms are stretched upwards, and she is standing with her feet shoulder-width apart, giving a sense of intimacy and closeness to the viewer. The background is a simple, minimalist bedroom with a white bed and a large window, allowing soft natural light to flood the space. The image is well-lit and has a soft, dreamy quality to it, making it a realistic and intimate scene. Additional descriptors: 34 years old, beautiful, curvy, sultry, luscious, seductive, busty, tantalizing, provocative, juicy, erotic, plump, tempting, racy, voluptuous, siren-like, smoldering, naughty."
"A photo-realistic shoot from a profile camera angle, featuring a young woman standing in a dimly lit room with a pink light casting a shadow behind her. The woman, who appears to be 20 years old, is standing in the center of the frame, facing the viewer with her eyes closed and her hands clasped together in front of her. She is wearing a black leather leotard, a high-waisted skirt, and thigh-high stockings, and has a slender physique. Her hair is styled in a bun, and she has a serious expression on her face. The background of the image is a simple, minimalist room with white walls and a pink gradient, creating a soft and ethereal atmosphere. The lighting is soft and diffused, casting a warm glow on the woman's skin, and the style is realistic with a focus on texture and detail."
"A photo-realistic scene captured as if recorded on an iPhone, featuring a nude 20-year-old woman with shoulder-length black hair walking toward the camera in a strip club setting. She has a curvy physique with jiggly thighs and a big ass, emphasized by her confident stride. The background includes a dimly lit strip club vibe with neon lights (purple and red hues), a reflective stage floor, and subtle smoke in the air, creating a sultry and immersive atmosphere. The lighting highlights her skin with a warm, seductive glow, and the style is realistic with a focus on motion and texture."
Ensure the WAN 2.1 models are correctly placed in ComfyUI’s directories as shown above.
Use negative prompts (e.g., "blurry, low detail, unnatural proportions") to polish outputs.
Adjust CFG scale (6 recommended) for prompt fidelity.
Experiment with LoRA weight (0.6–1.0) to control the intensity of the "PAWG" enhancement.
Dive into the WAN21 - PAWG LoRA and sculpt bold, luscious, and tempting figures that command attention with every curve!
Description
FAQ
Comments (19)
GYAT!
Gonna need a little more info, What is it meant to do?
It generates Phat Ass White Girls probably.
@funscripter627 😂😂😂😂
Can you include the prompts for the examples? I'm having trouble replicating them.
true
click on remix icon when over pic and it gives the prompt used....A woman is twerking, moving her hips and thighs in a rhythmic manner, rhythmic twerking emphasizing subtle physics-driven movements, fluid and smooth motion, she is looking back at the viewer, realistic
A woman is twerking, moving her hips and thighs in a rhythmic manner, rhythmic twerking emphasizing subtle physics-driven movements, fluid and smooth motion, she is looking back at the viewer, realistic
It's frustrating seeing this great content and having no idea how to recreate it.
If you have a powerful GPU, download Pinokio, then install WAN through it. Download the necessary models, and from there it's pretty self-explanatory.
@xxfortunadoxx888 Thanks for that info, really appreciated!!👍🏼
Is it at least minimally possible to use with a 6gb or 8gb GPU card, like GTX2060/2070?
After a week researching on this matter can answer you. On 2070 laptop 8 Gb card as mine, you can run quantized 14B models as Fusion TTV, VACE, probably Fusion ITV, didn't try yet. Googgle the QuantStack hugginface. You gonna need ComfyUI-GGUF custom node to load GGIF models. With Causvid working well. This is just entry level to get 14B loras works.
so good for adding jiggle
Would anyone be willing to share a minimal workflow that works with this LoRA to help get me started?
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.