English Bulldog Children's Book LoRA

This repository contains a LoRA (Low-Rank Adaptation) model trained to generate children's book-style illustrations of an English Bulldog character. The model was developed using a multi-stage process combining Deepseek's Janus Pro for base image generation and OmniControl for consistent character variations.

Generated with my lora

Demo

Try it out here: Hugging Face Spaces

Model Development Process

1. Base Character Generation

  • Used Deepseek's Janus Pro 7B model to create the initial base character
  • Focused on creating a watercolor-style English Bulldog with child-friendly features
  • Optimized for a consistent, appealing character design

2. Character Variations

Generated 100 variations using OmniControl with the following categories:

  • Basic Poses & Views (25 variations)
  • Expressions & Emotions (25 variations)
  • Activities & Scenarios (25 variations)
  • Seasonal & Special Themes (25 variations)

3. LoRA Training

  • Base Model: runwayml/stable-diffusion-v1-5
  • Training Parameters:
    • Resolution: 512x512
    • Learning Rate: 1e-4
    • Training Steps: 600
    • Scheduler: constant
    • No warmup steps

Usage

To use this LoRA for generating English Bulldog illustrations:

from diffusers import StableDiffusionPipeline
import torch

# Load the base model
pipe = StableDiffusionPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    torch_dtype=torch.float16
).to("cuda")

# Load the LoRA weights
pipe.load_lora_weights("Prof-Hunt/lora-bulldog")

# Generate an image
prompt = "A watercolor English Bulldog reading a storybook to children, child's drawing style, warm colors"
image = pipe(prompt, num_inference_steps=30, guidance_scale=7.5).images[0]
image.save("bulldog_illustration.png")

Examples

Here are some example prompts that work well with this LoRA:

  1. Basic Character Prompts:
"A watercolor English Bulldog, sitting, child's drawing style, front view, crayon texture"
"A watercolor English Bulldog, side profile, child's drawing style, wobbly lines, big eyes"
  1. Emotional Expressions:
"A watercolor English Bulldog, big smile, child's drawing style, happy colors"
"A watercolor English Bulldog, curious expression, child's drawing style, bright eyes"
  1. Activity Scenes:
"A watercolor English Bulldog reading a book, child's drawing style, glasses on, scholarly"
"A watercolor English Bulldog painting picture, child's drawing style, holding brush, artist dog"

Model Details

This LoRA was trained to maintain consistent character design while allowing for flexibility in poses, expressions, and activities. The watercolor style and child-friendly aesthetics make it particularly suitable for children's book illustrations.

Training Dataset Structure

  • Total Images: 100
  • Categories: 4 main categories with 25 variations each
  • Style: Consistent watercolor, child-friendly art style
  • Resolution: 512x512 pixels

License

This model is intended for creating children's book illustrations and similar family-friendly content. Please respect content guidelines and usage rights.

Acknowledgments

  • Deepseek AI for the Janus Pro model
  • OmniControl for variation generation capabilities
  • Hugging Face for infrastructure and model hosting

Citations

@misc{lora-bulldog,
    author = {Prof-Hunt},
    title = {English Bulldog Children's Book LoRA},
    year = {2024},
    publisher = {Hugging Face},
    journal = {Hugging Face Model Hub},
    howpublished = {\url{https://huggingface.co/Prof-Hunt/lora-bulldog}}
}
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and HF Inference API was unable to determine this model's library.

Spaces using Prof-Hunt/lora-bulldog 2