<h3 id="the-lora-works-well-with-base-or-sft-models-and-turbo-loras-and-was-trained-withoutmasked-faces.">The LoRa works well with Base or SFT Models and Turbo LoRas and was trained without/masked faces.</h3><p>Feel free to add your own character or style LoRas for example.</p><p></p><h3 id="for-best-results-use-slighly-lower-lora-weights:-around-0.6.">For best results use slighly lower LoRa weights: around 0.6.</h3><p><strong>You can try weights up to around 0.8 for a stronger effect, but it might produce missing/distorded limbs sometimes.</strong></p><p></p><h3 id="trigger-word-is:-owkx_swimsuit."><strong>Trigger word is: owkx_swimsuit.</strong></h3><p><strong>You might use additional trigger, like: one-piece, competition, neopren, translucent, etc.</strong></p><p></p><p>The LoRa sometimes seems to tend to generate asian woman. To avoid this use prompts like caucasian/european/american woman for example and/or specify hair color/hair style or just use an additional character LoRa.</p><p></p><p>You also might try the Kromoss checkpoint I used for my samples. It generates high quality photorealistic images with brilliant colors.</p><p></p><p>And as allways - have fun 🙂</p>

Files with SHA256: 4f2c4f9a6871d61926bd5d95c557fd45f737c299a8e9092d723099995b7cd6fb