Членство PixAI
PixAI
PixAI
Generate
Sign in
Головна
Генерація
Моделі
Studio
Конкурс
Mio.2

Conversational creation

Панель інструментів
Guide

User LoRA

XL
Інше

Aphrodisiac Mist / Lewd Atmosphere - V1

#Smoke#Aphrodisiac#Fog

Інструменти ШІ

  • OC Maker
  • AI Anime Generator
  • VTuber Maker
  • Sonic OC Maker
  • Hazbin Hotel OC Maker
  • Anime Character Creator
  • AI Figure Generator
  • Genshin OC Maker
  • All AI Tools

Огляд

  • Трендові теги
  • Рейтинг
  • Ринок Моделей
  • Конкурс
  • Новини

Про нас

  • Guide
  • Як користуватися PixAI
  • Tsubaki.2
  • Знайомство з Mio
  • Правила контенту

Ціни та довідка

  • Членство
  • Пакети кредитів
  • Контакти

Мобільний додаток

  • Завантажте застосунок PixAI
  • App Store
  • Google Play
© 2026 PixAI
  • Умови надання послуг
  • Політика конфіденційності
  • Політика авторського права
  • Attributions
  • Закон про визначені комерційні операції
Aphrodisiac Mist / Lewd Atmosphere - V1 - AI Model cover image
Завантажено
11 вер. 2025 р., 20:54
Використань
183
Відгуки
Немає відгуків
Тригерні слова

lovesmoke

Дозволи
  • Allow image generation & sharing
  • Дозволити користувачам завантажувати вашу модель
  • Комерційне використання

Опис

This LoRA either adds in a pink smokey/steamy effect to your image or modifies liquid and gasses to have a pink appearance. This was an experimental attempt at training the concept, and there is a lot of room for improvement in the LoRA. The aim was to replicate introducing the appearance strong smell/incense into the image output.The strength can be controlled through the weight of the LoRA and the weight of the trigger word "lovesmoke". Optimal weight is 1.0 ~ 1.2. Weights at or above 1.5 tended to ruin the output, so I don't recommend it in most cases. Lower weights work fine, but too low and you lose the effect entirely.The LoRA has a tendency to make the smoke very thick and appear as if it is being blown out, usually from a body part. This was not intentional, but can give some interesting outputs. If you don't want this it usually can be fixed by pulling the weights down, but it may just be something in your prompt causing it. I'm not entirely sure why the model trained this way. It is something I would like to fix in future versions. I had trouble collecting a dataset for exactly what I wanted, so I ended up trying to paint the effect onto images to build a dataset. Inconsistency there probably attributed to issues with the LoRA. On a positive I was able to use a diverse amount of images so the style impact was lessened as opposed to if I had just used the limited dataset I had gathered before then.

Коментарі