💥 Makino Kanna from Tamako Market

— She is the laid-back and deadpan one of Tamako’s group, usually moving at her own pace no matter what is happening around her. Being raised in a family of carpenters also rubbed off on her quite a lot, especially with her strange love for measurements, construction, and getting those precious right angles just right. She might look half-asleep most of the time, but her unexpectedly blunt comments and weird little habits make her surprisingly entertaining :)

>Tamako Market 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

🌳 Reducing wall-like details in outdoor scenes

The raw dataset has very little nature scenery and mostly consists of indoor locations or backgrounds with buildings.

Because of this, outdoor prompts can occasionally inherit wall-like structures even when the scene is supposed to be somewhere like a park or other natural environment.

If this happens, try using:
🚫 Negative prompt: wall, indoors

This will not completely remove the dataset bias, but it can help reduce those unwanted structures.

🎒 Preventing multiple school bags

Prompting her school bag can sometimes result in more than one bag appearing.

I recommend keeping the positive prompt simple:
✅ Positive prompt:school_bag
🚫 Negative prompt: multiple_bags, weapon_bag

Using only school_bag instead of stacking several bag-related tags seems to work better for me.

👗 Suppressing baked-in pleated skirt details

Her pleated skirt seems to be learned a little too strongly and can occasionally influence other outfits.

If pleated details keep appearing on clothes that should not have them, try using:
🚫 Negative prompt: pleated_skirt

You can also gradually lower the LoRA weight if the detail is still persistent.
I recommend trying the negative prompt first, since lowering the LoRA too much can also weaken her face, hair, and overall character identity.

🐰 Preventing the school emblem from leaking

Her uniform has a small bunny-like school logo, but the source coverage for it is not good enough for reliable fine details.
This version instead focuses more on preventing the emblem from appearing automatically on unrelated outfits.

If it still appears on another outfit, try using:
🚫 Negative prompt: school_emblem

The emblem itself may also look inconsistent when you intentionally prompt it, simply because it is a very small feature with limited dataset coverage.

⚖️ LoRA strength and outfit flexibility

The raw dataset almost exclusively shows her wearing the same school uniform with an open black cardigan, so completely different outfits may still need stronger prompting.

A weight of 1 can still work with different outfits, especially when the new outfit is clearly defined.

Only lower the LoRA weight gradually when default outfit details keep leaking even after using clear outfit tags and negative prompts. Lowering it too much may also weaken her face, hair, and overall character 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

🎬 Anime and movie screenshots with official art support

This LoRA was trained using a mix of anime screenshots, movie screenshots, and Megami Magazine scans. The movie screenshots received some color correction to keep them closer to the rest of the dataset, while some magazine images were adjusted to reduce unnecessary clutter around the character.

🏫 Limited environment variety

The raw dataset mostly consists of indoor scenes or backgrounds containing buildings and walls, with very little proper nature scenery. Because of this, some environment bias is still expected, especially when generating outdoor scenes.

👗 Default outfit source with sweater-free uniform support

This LoRA was trained entirely using source images where Kanna wears her default school outfit with the blue sweater, so some bias toward the sweater should be expected.

Even though the series never actually shows her wearing the pinafore dress without the sweater, this version was specifically optimized to still make that combination possible during generation.

🐰 School emblem prioritizes flexibility

Her uniform has a small bunny-like logo, but the feature has poor coverage and is too small to reproduce consistently.

Instead of focusing too much on accurately reproducing such a tiny detail, I prioritized preventing the emblem from becoming heavily baked into the character and appearing on every other outfit.

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