Remember to use the trigger word jeonheejin. Best results (or maybe not idk) come from using LORA weight 0.6-0.9 and Hires fix with latent bicubic antialiased, 20 steps, 0.5 denoising strength, and 1.25x upscale. I've used <a target="_blank" rel="ugc" href="https://civarchive.com/models/6424/chilloutmix">ChilloutMix</a> to generate images since the LORA is trained on Basil Mix.Version 2: This version is trained very differently on Hassanblend, works with LORA weight 1.0. I used <a target="_blank" rel="ugc" href="https://civarchive.com/models/17449/level4">Level4</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/9052/lofi">LOFI</a>, and <a target="_blank" rel="ugc" href="https://civarchive.com/models/20282/henmixreal">henmix_real</a> in addition to Chilloutmix. Could also mix this version with V1, e.g. "&lt;lora:heejinOTW_v2:0.5&gt; &lt;lora:heejinOTW_v1:0.35&gt;"

Heejin

LORA training settings are inspired by <a rel="ugc" href="https://civitai.com/user/alelele">Alelele </a>but I trained a much larger dataset on Basil Mix using a longer training time. This is why I recommend a weight of 0.6-0.9.

Files with SHA256: 482c469a37db3a752587c7aba39aacce002330b773933007d469811d835f3fa9