<h2 id="heading-5896">Disclaimer: This model is actually in [alpha]</h2>Due to the lack of images to train on, and the inconcistency in those, it is pretty hard atm to have decent outputs.This model is aimed to help in the composition of images with the tag "chest_jewel".It is a very nieche tag that, more often than not, is ignored by stable diffusion, and that generate pretty inconsistent output.<hr /><hr /><h2 id="heading-5897">[Alpha]:</h2>This model can help with the general composition, on the 60 generation i did as a test, the chest_jewel was always present and in the right spot.The issue is that it is often poorly detailed, or a low quality jewel that need some inpaint.The alpha version works with most of loras, and can be combined with different clothing attire to generate output near the desired ones (but often more similar to a brooch).It works with character loras, and poorly trained ones aswell.<h2 id="heading-5898">Current known issues:</h2><ul><li>The model tends to generate a more feminine character, even when you try to generate a male.</li><li>The model often generate a "magical girl" like character, or atleast it impacts on the generation of the outfit to make it more similar to that kind of outfit (this can be solved lowering the weight of the lora, but it makes the jewel less polished).</li><li>This model is heavely impacted by the tags concerning the xenoblade series, making the character often wearing and armored dress/armor.</li></ul>Actual Suggested settings:<pre><code>activation keyword: EmbeddedJewel
suggested tags: Chest_jewel, soul gem, jewels,[kind of gem you want
 in output, for example red_gemstone to have a ruby]
suggested weight: 0.6 - 1
suggested negative prompt: armor, armored dress, magical girl</code></pre>P.s. don't expect excellent output from this lora, i am sharing it just to increase the minimal quantity of images of this tag existing online, so that the training of the next version can be more consistent.

EmbeddedJewel

EmbeddedJewel.safetensors

alpha [experimental]

Disclaimer: This model is actually in [alpha]Due to the lack of images to train on, and the inconcistency in those, it is pretty hard atm to have decent outputs Actual Suggested settings:<ul><li>suggested weight: 0.6 - 1.0</li><li>activation keyword: EmbeddedJewel</li><li>suggested tags: Chest_jewel, soul gem, jewels,[kind of gem you want in output, for example red_gemstone to have a ruby]</li><li>suggested negative prompt: armor, armored dress, magical girl</li></ul>This model is aimed to help in the composition of images with the tag "chest_jewel".It is a very nieche tag that, more often than not, is ignored by stable diffusion, and that generate pretty inconsistent output

Files with SHA256: 539eaa035e6b207a6b8568e1e7edfa12df163704efd3a5824127902b02ff38d1