Sign In

A challenge to explore the full potential of my LoRA model

1,000 BUZZ
Awarded
7
3
1
20
Started: Aug 22, 2024
fluxlora test
Bounty Type
Image Creation
Started
Aug 22, 2024
Deadline
Sep 7, 2024

About this bounty

https://civitai.com/models/656320/korean-fashion

I’ve trained a LoRA using the Flux-dev model, utilizing 170 meticulously curated 4K high-resolution Korean fashion photo shoots. Each image has been carefully captioned in detail using the GPT-4o vision function, ensuring a high-quality dataset aimed at producing stunning results.


Caption Style

  • A young woman with shoulder-length dark hair sits in a modern, sunlit living room on a black leather chair. She is dressed in a sleek navy blue suit with a sleeveless black top, exuding a chic and contemporary style. The setting is warm and inviting, with sunlight streaming through the window, casting soft shadows on the floor and across the white textured rug. A tall, stylish floor lamp stands beside her, adding to the elegant atmosphere. The woman appears deep in thought, her hands slightly gesturing, reflecting a moment of introspection in a serene, well-appointed space.

  • A young woman with short, dark hair sits on a wooden park bench during a sunny day. She wears a textured, olive green sweater with oversized sleeves, resting her chin on her hand in a thoughtful pose. The background reveals an urban setting with tall, leafless trees and residential buildings, reflecting a calm, early spring atmosphere. A soft, golden light illuminates her face, casting gentle shadows that enhance the warm and relaxed mood of the scene. To the side, a glimpse of a takeaway coffee cup adds a casual element to the composition.

  • ...


After multiple attempts and failures, spending countless Buzz in the process, I finally crafted a LoRA that not only captures the training data remarkably well but also enhances the overall performance of the Flux model, surpassing even my own expectations.

However, testing this LoRA on my own has its limits and has become exhausting, which is why I’ve decided to launch this bounty. I’m offering 1000 Buzz to the creator of the most impressive image generated using only the Flux-dev model and my LoRA. The winner will be chosen based on a mix of emoji reactions and my personal judgment. Thank you!

Entries

Discussion