my-face-lora

Jojo LoRA - DreamBooth Fine-Tuned Stable Diffusion v1.5

Model Overview

This is a LoRA fine-tuned Stable Diffusion v1.5 model, trained with DreamBooth on images of Jojo (myself 😁).
The model allows generating high-quality personalized images when prompted with the keyword "jojo".

• Base Model: runwayml/stable-diffusion-v1-5
• Technique: Low-Rank Adaptation (LoRA)
• Training Method: DreamBooth with prior preservation
• Resolution: 512 × 512
• Steps: 800
• Rank: 4

---

How to Use

Load this LoRA with the Stable Diffusion pipeline from 🤗 Diffusers:

from diffusers import StableDiffusionPipeline
import torch

model_id = "runwayml/stable-diffusion-v1-5"
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda")

# Load LoRA weights
pipe.unet.load_attn_procs("Swarnadeep-28/my-face-lora/jojo_lora")

# Prompt
prompt = "photo of jojo person in cinematic lighting"
image = pipe(prompt).images[0]
image.save("jojo_output.png")

Example Prompts

photo of jojo person smiling

portrait of jojo person wearing sunglasses

cinematic close-up of jojo person

Training Details

• Instance Prompt: photo of jojo person

• Class Prompt: photo of a person

• Batch Size: 1

• Learning Rate: 1e-4

• Scheduler: constant

Limitations

May generate artifacts in complex scenes.

Results depend on prompt quality.

Trained on limited dataset (Jojo face).