Capture the guiding light of the coast with this new LoRA, meticulously trained on a curated collection of photographic-quality lighthouse images. This model excels at rendering these iconic beacons with stunning detail and realism. Whether you're crafting a dramatic seascape lashed by storms or a tranquil coastal scene bathed in a golden sunset, this LoRA provides the flexibility to generate the perfect lighthouse for your artwork. You can add a powerful touch of nautical majesty to your creations. <h1 id="lighthouse-xl-v2.0">Lighthouse XL v2.0</h1>Version 2.0 of the Lighthouse XL LoRA builds upon the successful v0.9 proof-of-concept by expanding the image set for training, with a larger dataset derived from a curated set of images using AI and my <a target="_blank" rel="ugc" href="https://github.com/AlbertJBurton/LoRA_Dataset_Curator">Lora Dataset Curator</a>. This evaluation image set comprised over 500 high-quality photographs of lighthouses. I chose 250 of the top-scoring pictures that illustrate the concept of a lighthouse. This improved and optimized training set greatly enhances the ability of any SDXL checkpoint to generate a high-quality image of a lighthouse. I tested the LoRA with several popular SXDL fine-tuned checkpoints with great results (see below).<h2 id="notes-on-training">Notes on Training</h2>I trained the original v0.9 on 100 images, mostly from my personal collection of photographs I have taken over the years. A large majority of those images were taken on black-and-white film, for which the scanning process picked up the film grain. Sometimes, to the detriment of the image. I tried several techniques to minimize the impact of film grain on the images produced by v0.9, but it still appears as banding in some generated images. Version 2.0 attempts to address the issues in v0.9, beginning with a larger set of training images, specifically replacing the film images that did not scan well. I added images to my collection until I had 500+ quality photographs of lighthouses in diverse environments and lighting conditions. These images were fed to my LoRA Dataset Curator software, which scored each image based on how well it aligns with the concept of "lighthouse." I rejected all images that did not score the maximum value (1.0). This resulted in a set of approximately 375 curated images. From this set, I culled what I deemed to be duplicates based on how they interpreted the concept and settled on a set of 250 training images. The curation process yielded a superior training image set, ultimately resulting in a more effective LoRA that generates high-quality images under diverse conditions. Your feedback, whether positive or negative, is crucial in determining whether my belief is correct. If you are interested, you can download my <a target="_blank" rel="ugc" href="https://github.com/AlbertJBurton/LoRA_Dataset_Curator">LoRA Dataset Curator</a> software and run it on your system. It is entirely free and may help you curate your own concept LoRAs, where the interpretation of the concept is crucial. Using AI to analyze the image helps mitigate some of the biases and subjectivity that can creep in when evaluating images for a particular concept.<h2 id="checkpoints-tested">Checkpoints Tested</h2>I tested the following checkpoints with the Lighthouse XL v2.0 LoRA &lt;strength:1.0&gt; with the prompt and settings listed below. Each image contains the ComfyUI workflow that generated the image. Drag these images into your ComfyUI environment to reproduce the simple workflow I used. The images you see were not subjected to any refining, sharpening, or HiRes fixing. They were downscaled by a factor of 0.5 to a convenient size.You can find links to the checkpoints listed below in the Suggested Resources section. Prompt: A red and white banded lighthouse sits at the end of a rock wall extending 200 feet from the shore, during sunset. Large waves are crashing against the rocky coast. Seed: 0 CFG: 7.0 Sampler: Euler Scheduler: beta <h3 id="checkpoint:-sdxl-base-1.0">Checkpoint: SDXL Base 1.0</h3><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/471bac73-9822-4286-9522-c9f4f20c224a/width=525/471bac73-9822-4286-9522-c9f4f20c224a.jpeg" /><h3 id="checkpoint:-juggernaut-xl-ragnarok">Checkpoint: Juggernaut XL Ragnarok</h3><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/10931221-38a6-4c59-8403-80ec6e6e8893/width=525/10931221-38a6-4c59-8403-80ec6e6e8893.jpeg" /><h3 id="checkpoint:-hello-world-xl-v7.0">Checkpoint: Hello World XL v7.0 <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1b3fcb8a-4789-4bfc-843b-f7dcd1234835/width=525/1b3fcb8a-4789-4bfc-843b-f7dcd1234835.jpeg" /></h3><h3 id="checkpoint:-cyberrealistic-xl-v7.0">Checkpoint: Cyberrealistic XL v7.0 <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fe9589d2-dd84-405b-a28d-b51bc68bf399/width=525/fe9589d2-dd84-405b-a28d-b51bc68bf399.jpeg" /></h3><h3 id="checkpoint:-landscape-realistic-alpha-1.0">Checkpoint: Landscape Realistic Alpha 1.0</h3><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/31b012bb-191c-4351-801c-c9c383d6a525/width=525/31b012bb-191c-4351-801c-c9c383d6a525.jpeg" /><h3 id="notes-on-the-test-images">Notes on the Test Images</h3>I generated these images on an AMD "Strix Halo" machine with an AMD Ryzen AI Max+ 396 CPU and an integrated Radeon 8060S GPU. Your mileage may vary. In other words, you may never exactly reproduce these images on your system. Welcome to the world of random noise diffusion.<h2 id="restrictions-and-limitations">Restrictions and Limitations</h2>I removed the restrictions on generation that I placed on v0.9. I imposed those restrictions simply because there was a possibility that the proof-of-concept was faulty, and I did not want people generating bad images with a faulty LoRA. I encourage you to create a prompt to generate your lighthouse masterpiece, print it large, and hang it on your wall for everyone to admire. However, using this LoRA and diffusion models to create images for sale is outside the permitted use of this LoRA. If you would like to train your own LoRAs and generate images for sale, then by all means go for it. However, using other people's effort and expense to enrich yourself is something you will have to atone for later in life. That applies to all models and LoRAs, not just mine.<h2 id="revision-history">Revision History</h2><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/39e90295-ab6c-493d-bec6-d4ea1f598a67/width=525/39e90295-ab6c-493d-bec6-d4ea1f598a67.jpeg" />

Lighthouse XL

Initial version trained with one hundred curated images of lighthouses and associated buildings.

Lighthouse_XL.safetensors

V0.9

v2.0

Lighthouse XL v2.0Version 2.0 of the Lighthouse XL LoRA builds upon the successful version 0.9 proof-of-concept by expanding the image set for training with a larger dataset derived from a curated set of images using AI and my <a target="_blank" rel="ugc" href="https://github.com/AlbertJBurton/LoRA_Dataset_Curator">Lora Dataset Curator.</a> This evaluation included over 500 photographic-quality images of lighthouses. I chose 250 of the top-scoring images that illustrate the concept of a lighthouse. This improved and optimized training set greatly enhances the ability of any SDXL checkpoint to generate a high-quality image of a lighthouse. I tested the LoRA with several popular SXDL fine-tuned checkpoints with great results. See the v2.0 description for the models we tested and example results.I have also flipped off the restrictions on generation that I placed on v0.9. I imposed those restrictions simply because there was a possibility that the proof-of-concept was faulty, and I did not want people generating bad images with a faulty LoRA.Your feedback is welcome, positive or negative. I use all of the feedback to improve or fix issues that I have missed in my testing.FWIW, I am currently evaluating the native ability of FLUX.1 Dev to generate images of a lighthouse. If I believe I can enhance the generated images from FLUX, I’ll consider creating a version of this LoRA for FLUX.

Files with SHA256: 6242b8e06807102349b4c1de41705307d29a1c51827f54afda9a8866a8e1c13d