<h1 id="heading-874">Eonsang style for Pony Diffusion XL</h1><p>This is a LoRA to replicate the style of <a target="_blank" rel="ugc" href="https://www.pixiv.net/en/users/24600840">Eonsang</a>.</p><p>The results are most likely not optimal - these are my first experiments with PDXL. Expect better stuff!</p><p></p><p></p><h2 id="heading-56"><strong><span style="color:rgb(253, 126, 20)">Version 1.0a</span> <span style="color:rgb(250, 82, 82)">NEW</span></strong></h2><h3 id="heading-57"><strong>Weight</strong></h3><p>A weight value of <strong>1 </strong>is totally fine.</p><p></p><h3 id="heading-58"><strong>Trigger words</strong></h3><p>No trigger words.</p><p></p><h3 id="heading-59"><strong>Known issues</strong></h3><p>You may need to lower the weight in certain situations. I would also suggest putting 'watermark' in the negative prompt.</p><p></p><p></p><h2 id="heading-148"><strong><span style="color:rgb(253, 126, 20)">Version 0.2</span></strong></h2><h3 id="heading-149"><strong>Weight</strong></h3><p>A weight value of <strong>1 </strong>is totally fine.</p><p></p><h3 id="heading-150"><strong>Trigger words</strong></h3><p>No trigger words.</p><p></p><h3 id="heading-151"><strong>Known issues</strong></h3><p>It works very well with NSFW stuff (and this is not surprising), so if you use NSFW keywords in your prompt, you can expect something close to the artist's style.</p><p>If you DON'T use NSFW keywords, it doesn't work particularly well.</p>

Eonsang Style [PDXL]

<p>This is <strong>version 1.0a</strong>: I've used a rather <em>small</em> dataset for training (around 120 pictures) in which the artist's most recent works are included. A different training process has been used (compared to version 0.2).</p>

Files with SHA256: cd9110b262cf1b3ddd198d9b484de2211d9ec68b34774ed9d0e1494f5bc6a10a