alphaduriendur/avatar-weights-01

Flux.1-dev LoRA — Avatar

LoRA fine-tuned on Flux.1-dev for personalized avatar generation.

Training details

• Base model: black-forest-labs/FLUX.1-dev
• Instance prompt: a photo of alphaduriendur
• LoRA rank: 16
• Training steps: 1000
• Learning rate: 0.0004
• Resolution: 1024

Usage

from diffusers import FluxPipeline
import torch

pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev",
    torch_dtype=torch.bfloat16
)
pipe.load_lora_weights("alphaduriendur/avatar-weights-01")
pipe = pipe.to("cuda")

image = pipe(
    prompt="a photo of alphaduriendur, professional headshot, studio lighting, 8k",
    num_inference_steps=28,
    guidance_scale=3.5
).images[0]
image.save("avatar.png")