V4.5作品
▾
>

sword girls

2,355 Danbooru 投稿 · 最高分 208 · V4.5

seed 1234點擊圖片放大
NovelAI V4.5 使用「sword girls」標籤的 AI 生成範例圖
V4.5NAI Diffusion Anime V4.5
Promptsword girls, 1girl, solo, best quality, masterpiece, cowboy shot, sfw, fully clothed, looking at viewer
角色CHARACTERS IN SWORD_GIRLS · 4
[1/1]
cinia pacifica
CHAR0x0000
0
iri flina
CHAR0x0001
0
sita vilosa
CHAR0x0002
0
luthica preventer
CHAR0x0003
0

關於「sword girls」標籤

sword girls 是 NovelAI Diffusion 可以使用的作品標籤。Danbooru 上有 2,355 篇投稿,靜態圖的最高分數為 208。上方範例把這個標籤放在固定基準提示詞的開頭,並以 seed 1234 生成,因此與其他標籤的差異都來自這個標籤。

各模型畫出的 sword girls

V5

忠實遵循構圖與表情的指定,標籤效果呈現得較整齊。

V4.5

對標籤的反應更強,效果更容易看出來。

ILLUST

本機模型 WAI illustrious SDXL v17。可以看到同一標籤在 NovelAI 以外的表現。

sword girls 在提示詞中的寫法

先寫主體,再寫這個標籤,最後接品質標籤。

基本sword girls, solo, best quality, masterpiece
加繪師wlop, sword girls, solo

負面提示詞與完整設定請見 使用條款。

範例的生成條件

所有圖片都以 832×1216、seed 1234、23 step、k_euler_ancestral 生成。基準提示詞與負面提示詞對所有標籤相同,只替換要測試的標籤。被 NSFW 檢查判定的圖片會換 seed 重新生成或隱藏。

Danbooru wiki 摘錄

展開 wiki 內文 ▾

A Korean Flash-based online collectible card game developed by Zeonix, featuring over a thousand individual cards with which players may pick and play from in either a PvP or PvE setting.

New cards are regularly added on a sporadic basis. These cards are designed by artists from South Korea and Japan. Their lead artist is Salt (salty).

Game closed down due that the company managing it Zeonix went bankrupt, and the game's servers closed down permanently. Outside of a few unofficial PC ports, the game can't be played anymore.

On the 18th of August, 2020, Korean game company Game Meca announced they bought the rights to the game, with plans of rereleasing the game as a mobile game in early 2021. The mobile game version of Sword Girls was finally released in English in 2023 under the name Kana Tales.

Characters

Vita
Academy
Crux
Darklore
Neutral

External links

來源:Danbooru wiki · sword_girls ↗

常見問題

NovelAI 能辨識 sword girls 嗎?

請對照本頁的範例。本站所有標籤都使用相同的 seed 與基準提示詞,畫面上的差異都來自 sword girls。

sword girls 在提示詞中怎麼寫?

把「sword girls」寫在提示詞前段,並用逗號與其他標籤分開。

V5 和 V4.5 該用哪個?

V4.5 更能表現標籤的特色,V5 的姿勢更穩定、破綻更少。