Nana Osaki · 大崎娜娜 (2006 NANA anime)

大崎娜娜 —— 2006 年动画版本，我的最爱，最近灵光一闪，有了这个 LoRA。

触发词：osaki_nana

【版本】

v1.0 SFW：角色本体，日常出图用这张。

v1.1 NSFW：成人版，解锁层已经融进同一个文件里。

【用法】

1. 触发词放在提示词开头。她的发型、五官、妆容都焊在触发词里，不需要再写外观描述（写了反而会跟触发词抢权重）。

2. 想要 2006 电视动画赛璐珞质感，在提示词末尾加上：

Render everything in a highly atmospheric and cinematic anime style, 2000s Japanese cel animation with clean crisp ink lines and cel shading.

3. 想要「2D 人物走进真实世界」（真实照片背景 + 平涂人物），再加上：

Draw the character as a flat 2D cel animation figure with clean linework and cel shading, while the background stays a real photograph.

4. 提示词较长、或人物在画面里较小时，建议补一句妆容特征：

heavy smoky eyeliner with long lashes and clearly drawn rose-red lips

【训练】

NSFW 版 = 本模型 ×1.1 与 Mystic XXX v3 ×0.8 融合后截断至 rank 32。（如有侵权可告知撤销）

【已知限制】

· 人物在画面里越小越不像（全身/大远景时脸部细节会简化）。

· 手部容易崩：写清两只手分别拿什么、按在哪里，可以帮助改善。

Nana Osaki — her 2006 anime version, and my all-time favorite. A recent spark of inspiration led me to create this LoRA.

Trigger word: osaki_nana

【Versions】
v1.0 SFW: The core character LoRA for everyday image generation.
v1.1 NSFW: The adult version, with the unlock layer already merged into a single file.

【How to Use】

1. Place the trigger word at the beginning of your prompt. Her hairstyle, facial features, and makeup are already tied to it, so there is no need to describe her appearance separately—doing so may compete with the trigger word.

2. For the cel-shaded look of a 2006 TV anime, append:
   Render everything in a highly atmospheric and cinematic anime style, 2000s Japanese cel animation with clean crisp ink lines and cel shading.

3. For a “2D character stepping into the real world” look—a photographic background with a flat-shaded character—also append:
   Draw the character as a flat 2D cel animation figure with clean linework and cel shading, while the background stays a real photograph.

4. If your prompt is long or the character appears small in the frame, try adding these makeup details:
   heavy smoky eyeliner with long lashes and clearly drawn rose-red lips

【Training】
The NSFW version combines this model at ×1.1 with Mystic XXX v3 at ×0.8, then truncates the result to rank 32. (If this infringes your rights, please let me know and I will take it down.)

【Known Limitations】
· Character likeness becomes weaker as she gets smaller in the frame; facial details may simplify in full-body or distant shots.
· Hands can be inconsistent. Clearly specifying what each hand is holding or where it is placed can help.