✨ Kon from OtakuVS

— She is one of those characters who stands out by being the closest thing to “normal” in a world full of chaos. Compared to the others, she has a more grounded and organized vibe, often feeling like the one trying to keep everything from completely falling apart. That contrast is exactly what makes her fun, since her calmer presence bounces so well off the nonsense happening around her. Even with her limited screen time, she still manages to feel memorable, charming, and very much like more than just a background face.

It's intentionally called as "season 2" when there's no season 1 version. In season 1, Kon is not appeared yet and I find putting season 2 in the name is better to tell you about what kind of style you should expect since through season, OtakuVS have changed their style a bit drasticly.

>OtakuVS COLLECTION<

HERE<

🖼️ How I Generate My Images

I made a full article that explains how I generate my images — including my settings, prompting tips, and even my potato PC setup 😅

👉 Check it out here if you’re curious or want to try the same method.

🔞 NSFW gallery (18+)

Check my account here to see the spicy showcase that I can't show here, if any.

🎨 Prompt Tips

These are optional starting points for handling a few behaviors of this version. You may still need some retries depending on the checkpoint, prompt, and LoRA strength.

👕 Handling unwanted shirt pockets

The shirt pocket was intentionally removed from the training dataset because its appearance was inconsistent in the original source.

If a checkpoint still introduces a pocket on its own, you can try:
🚫 Negative prompt: pocket

🧥 Default outfit details appearing on other outfits

When a default outfit detail keeps appearing in a different outfit, add the exact leaking detail to the negative prompt.

For example :
🚫 Negative prompt: name_tag

Use the specific leaking tag instead of adding many unrelated negative tags.

⚖️ LoRA strength and outfit flexibility

The dataset is still biased toward the character’s usual outfit, so completely different outfits may need clearer prompting to work reliably.

A LoRA weight of 1 can still work well for alternate outfits, especially when the new clothing is clearly defined in the prompt.

Only lower the LoRA weight gradually if default outfit details keep leaking even after using clear outfit tags and negative prompts. Lowering it too much can also weaken the character’s face, hair, and overall identity.

📊 Lora Project

If you want to know what characters that I'm covering next you can check these link :
👉 Personal Project
👉 Commision Project

If this model sparked something in you, show some love with a like, review, or buzz.
Now go—

Be wild

Description

🎬 Trained from the full YouTube video series

This LoRA was trained using screenshots covering her appearances throughout the YouTube video series, with some personal adjustments made to the dataset where needed.

The goal was to make use of as much of her available source material as possible while keeping the dataset consistent enough for character training.

👗 Supermarket uniform with green jacket variation

The source material only shows her wearing the supermarket employee uniform, so the dataset is naturally biased toward that outfit.

It also includes a variation where she wears the green jacket over the uniform, allowing both versions to be represented instead of treating the jacket as a permanent part of her appearance.

👕 Shirt pocket intentionally removed

Her original shirt sometimes appears with a chest pocket and sometimes without one, making the detail inconsistent across the source material.

Rather than trying to reproduce an unreliable detail, I removed the pocket from the training images where necessary and prioritized a more consistent shirt design instead.

As usual, the final result can still depend on your checkpoint, prompt, LoRA strength, CFG scale, camera angle, and generation settings.

LORA

Illustrious

by zetsubou

Download (Beta) View on CivitAI