anatomical deformities
The portraits generated by this model are more likely to exhibit anatomical deformities compared to those from flux.dev. I would like to ask what causes this issue—could it be due to the long training time?
Could you post testing prompts and results?
The left side is the flux.dev model, and the right side is the awportrait model, using the same seed and a guidance scale of 2.5, implemented with the diffusers inference framework.
prompt: a beautiful female model,age 26-40, ((Asian Face)), (((White complexion))), (thin face), (vintage long wavy hair), (pink_hair), laughing, , , , (interior, (next to a bed), natural light, bright day), photography, masterpiece, best quality, 8K, HDR, highres
a beautiful female model,age 26-40, ((Asian Face)), (((White complexion))), (thin face), (long wavy hair), (brown_hair), little smile, wear T-shirt, wearing jeans, wear Sports shoes, (curtain as background), photography, masterpiece, best quality, 8K, HDR, highres
Hi, I have tested the weights using the same prompts with same setting. Please find attached the comparison.
Prompt 1:
a beautiful female model,age 26-40, ((Asian Face)), (((White complexion))), (thin face), (vintage long wavy hair), (pink_hair), laughing, , , , (interior, (next to a bed), natural light, bright day), photography, masterpiece, best quality, 8K, HDR, highres
Org:
This repo:
Prompt 2:
a beautiful female model,age 26-40, ((Asian Face)), (((White complexion))), (thin face), (long wavy hair), (brown_hair), little smile, wear T-shirt, wearing jeans, wear Sports shoes, (curtain as background), photography, masterpiece, best quality, 8K, HDR, highres
Org:
This repo:
Would the deformity issues come from randomness? Or simply changing the prompts to coherent natural languages would probably solve it.
Are you using the ComfyUI workflow or the Diffusers pipeline? and lora merge weight == 0.9？
Is your inference code like the one below?
import torch
from diffusers import FluxPipeline
pipe = FluxPipeline.from_pretrained("Shakker-Labs/AWPortrait-FL", torch_dtype=torch.bfloat16)
pipe.to("cuda")
prompt = "close up portrait, Amidst the interplay of light and shadows in a photography studio,a soft spotlight traces the contours of a face,highlighting a figure clad in a sleek black turtleneck. The garment,hugging the skin with subtle luxury,complements the Caucasian model's understated makeup,embodying minimalist elegance. Behind,a pale gray backdrop extends,its fine texture shimmering subtly in the dim light,artfully balancing the composition and focusing attention on the subject. In a palette of black,gray,and skin tones,simplicity intertwines with profundity,as every detail whispers untold stories."
image = pipe(prompt,
num_inference_steps=24,
guidance_scale=3.5,
width=768, height=1024,
).images[0]
image.save(f"example.png")
Likewise. Permanent deformations. I use ComfyUI and checkpoint AWPortrait-FL.safetensors