A lightweight LoRA (Low-Rank Adaptation) adapter fine-tuned on the marvy/book-covers dataset to generate professional, publication-ready book covers using Stable Diffusion 1.5. Optimized for centered compositions, title-safe zones, and the mdjrny-v4 aesthetic style.

Model Details

Model Description

This model is a PEFT-compatible LoRA adapter that modifies the UNet attention layers of Stable Diffusion v1-5 to specialize in generating book cover artwork. It does not replace the base model — it must be loaded alongside runwayml/stable-diffusion-v1-5.

Key characteristics:

🎨 Trained to produce clean, centered compositions with space for title/author text
🚀 Lightweight (~20-50MB) vs. full model (~4GB)
🔁 Plug-and-play with any SD 1.5 pipeline via diffusers.load_lora_weights()
🎯 Style-consistent outputs using mdjrny-v4 prompting conventions
Developed by: [Manohar.G/ MSM]
Model type: LoRA Adapter (PEFT) for Text-to-Image Diffusion
Language(s): English (prompt conditioning)
License: OpenRAIL++
Finetuned from model: runwayml/stable-diffusion-v1-5

Model Sources

Repository: https://huggingface.co/manoaad/book-cover-lora-v1-f
Base Model: https://huggingface.co/runwayml/stable-diffusion-v1-5
Training Dataset: https://huggingface.co/datasets/marvy/book-covers
Demo [optional]: Coming soon — open to community contributions!

Uses

Direct Use

This LoRA adapter is intended for:

Authors, self-publishers, and designers generating custom book cover concepts
Creative prototyping for publishing workflows
Educational exploration of parameter-efficient fine-tuning for diffusion models

Example workflow:

python

pipe.load_lora_weights("manoaad/book-cover-lora-v1-f")
image = pipe("mdjrny-v4 style, book cover, mystery novel, foggy lighthouse").images[0]