<h1 id="logolabs.org-lite-lora-collection"><a target="_blank" rel="ugc" href="http://LogoLabs.org">LogoLabs.org</a> LITE - Lora Collection</h1>The following is a collection of our old prototype generation engine, which we use in the pipeline at LogoLabs.As part of my appreciation to the open-weight community, I've decided to open-weight all of the LoRA trains that were notable.<h2 id="lora-versions:">LoRA Versions:</h2><ol><li>0.0.1 (Flux 1 Dev): The earliest version, was trained on 1.000 sample curated logos, and represents the first version of the model.</li><li>0.1.0 (Flux 1 Dev): The second version represents a 2-epoch train on a dataset of 10.000 logos. This also includes adding style tags for better convergence, as well as improving the curation system.</li><li>1.0.0 (Coming Soon for Flux 2 Klein 9B Base): Will include better captioning, more varied amount of logos.</li><li>2.0.0 (In the data curation stage): We managed to curate 200.000 high quality logos from the most popular brands in the world, at very high quality (SVG), and we are thinking how to make this into a great logo creator for you to enjoy.</li></ol><h2 id="open-weight-plan:">Open Weight Plan:</h2>The plan is to provide this model as an open source alternative to our paid services, which is more intended for small businesses that need a quick logo, as well as support and money back guarantee.As I'm going to launch the project (<a target="_blank" rel="ugc" href="http://logolabs.org">logolabs.org</a>) and get customers to offset the costs, I'm going to release more varied LoRAs for multiple styles, to make it accessible to create a logo for your business for cheap, with hosted inference on our end.Unfortunately, as AI costs are high, we will keep the latest &amp; greatest models on our paid offering, but we hope to provide affordable (DIY) good logos via models that can fit on your video card, like Z-Image or Flux Klein 9B Base.<hr />For any questions, office@incorpo.ro

LogoLabs.org LITE - Logo Generation LoRA Collection from 50K+ Curated Logo Dataset

A more modern train of the initial early version.

Files with SHA256: 14dae27be29901811d0204d455afd3495b6291a594e9e7212df43f46c3f280d1