<a target="_blank" rel="ugc" href="https://patreon.com/psoft"><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/8dcbc56a-15e9-429e-9b36-c2bc22594678/width=525/8dcbc56a-15e9-429e-9b36-c2bc22594678.jpeg" /></a> Support my work on <a target="_blank" rel="ugc" href="https://www.patreon.com/psoft">Patreon</a> and <a target="_blank" rel="ugc" href="https://ko-fi.com/psoft">Ko-Fi</a> and get access to tutorials and exclusive models. Rate and leave a like if you enjoyed it, and follow for new models!<hr /><h3 id="heading-6">UPDATE: removed noise offset, slightly less training, better for combining with multiple other LoRAs</h3>Finally, after much trial and error, I believe this LoRA came out really well. But you be the judge: let me know what you think and post your own generated images. See features, how to use and examples below.New features:<ul><li>easy to make good gens on anime models like AOM2 (click "view more" in the gallery to see non cherry-picked batches)</li><li>compatible with other LoRAs, to fit your style, pose and even use it as an outfit/cosplay for other characters, since you can change hair color</li><li>works with various models, including anime, 2.5D and realistic (some realistic models may require more tweaking the prompt, weight and CFG to get the best results; see my examples)</li></ul>How to use:<ul><li>Works with weight <code>0.6-1.0</code> and CFG <code>5-8</code> , depending on the model (e.g. higher weight and CFG for realistic models)</li><li>Main trigger word: <code>shimakaze</code></li></ul><ul><li>The main trigger word will generally provide the outfit, but these trigger words can be used to emphasize the character/outfit: <code>blonde hair, striped thighhighs, microskirt, elbow gloves, sailor collar, anchor hair ornament, highleg panties, black panties</code></li><li>Other keywords associated with the character and context: <code>robot, turret, lifebuoy</code></li></ul>Additional infoI make these models for fun and for the challenge and release them for free for everyone to enjoy, for non-commercial use (see license). The one thing I ask of you is to provide rating and feedback! That helps me understand where I can improve and helps me stay motivated to produce more content!I accept suggestions or requests here or on Civitai discord server.You can also check my more polished images and upcoming models on my <a target="_blank" rel="ugc" href="https://www.pixiv.net/en/users/25545070">pixiv </a>profile.Batch grid (same prompts, different seeds, various models), AOM2<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/aa51b479-54a7-47f0-a94a-4eadeafb7300/width=525/aa51b479-54a7-47f0-a94a-4eadeafb7300" /><img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/195cf81f-b768-47db-13aa-6f555a094f00/width=525/195cf81f-b768-47db-13aa-6f555a094f00" />Counterfeit-v2.5<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4b6d438c-6418-465e-b136-37ef0f29d300/width=525/4b6d438c-6418-465e-b136-37ef0f29d300" /><img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/3c586907-ac29-48cd-80c3-d781dad9d100/width=525/3c586907-ac29-48cd-80c3-d781dad9d100" />

Shimakaze (kancolle) - Character

Trained on NAI with kohya-gui614 images (not cropped)Tagged WD1.4 with added trigger word and some edits.10 repeats, 20 epochs, batch size 6LR 0.0001, constant, Adam8bitresolution 512,512dim/alpha 8noise offset 0.05shuffle caption, keep tokens 1~2h runtime on rtx 4090

Files with SHA256: 30fd963248cf15420635c49e023bf56b8c6a819592c2c37cffa2969118b44315