<p><a target="_blank" rel="ugc" href="https://www.instagram.com/george_ahhh/">Who is Georgia?</a></p><h1 id="heading-1567">Usage Notes</h1><p>Georgia often has a nose ring in, and you'll sometimes see artifacts on her nose. You can put <code>(nose ring:1.3)</code> in the negative prompt to help deal with that.</p><h1 id="heading-1568">Versions</h1><p>Georgia was one of my earliest attempts at making a lora. In <code>v3</code> I put together some of my learnings to improve it, but her eyes still weren't very accurate. With <code>v4 locon</code>, I applied a more refined understanding of the same strategies I used with v3, as well as improving the captions. It's also a locon rather than a lora, which essentially makes the file smaller while retaining the quality characteristics of a larger lora.</p>

Georgia Ellenwood

<p>I found myself displeased with Georgia's eyes in previous versions of this lora.  I applied some of the things I've learned since then, and created a superior version.</p><p>The v3 was an early revamp using a split weight strategy.  But at that point I didn't have as clear a conception of what I wanted.  Now I'm hitting a 2:1 primary to supporting ratio.</p><p>I also updated the captions.  That didn't help with the artifacts her nose ring causes, however.</p>

sd15_GeorgiaEllenwood_locon_24_v1.safetensors

georgia.ellenwood.zip

v4 locon

&lt;p&gt;I found myself displeased with Georgia&#39;s eyes in previous versions of this lora.  I applied some of the things I&#39;ve learned since then, and created a superior version.&lt;/p&gt;&lt;p&gt;The v3 was an early revamp using a split weight strategy.  But at that point I didn&#39;t have as clear a conception of what I wanted.  Now I&#39;m hitting a 2:1 primary to supporting ratio.&lt;/p&gt;&lt;p&gt;I also updated the captions.  That didn&#39;t help with the artifacts her nose ring causes, however.&lt;/p&gt;

Files with SHA256: 5d45379617a7f6f50940e85286fe48a20729983e43fe55ca8f0f696979a2c2b6