Firefly (StellaSora) LoRA

This is a dedicated character LoRA trained for Firefly (StellaSora). It supports multiple original outfits and weapon accessories, and has been thoroughly stress-tested for outfit flexibility (clothing swaps).

🔑 Core Trigger Words

• Base Character: firefly_(stellasora)

• Daily Explorer Outfit: stella_explorer_(costume), blue_scarf, small pink camera

• Sherlock Penguin Suit: sherlock_(penguin_costume), penguin_pajamas

• Early White Nightgown: sora_pajamas_(nightgown), white_nightshirt

• Avalanche Weapon (Funnels): avalanche_(weapon), floating_weapon, funnel_(weapon)

💡 Suggested Prompt Formats

You can copy and paste the following base formats for your generations:

Daily Explorer Outfit

Plaintext

1girl, firefly (stellasora), stella explorer (costume), blue scarf, short hair, light brown hair, red eyes, pink small camera on belt,

Sherlock Penguin Suit

Plaintext

1girl, firefly (stellasora), sherlock (penguin costume), penguin pajamas, ahoge, blue scarf, red eyes, bare legs,

Early White Nightgown

Plaintext

1girl, firefly_(stellasora), sora_pajamas_(nightgown), white_nightshirt, oversized_shirt, short hair, light brown hair, wide_sleeves, long_sleeves, barefoot, white choker,

Equipping Weapon Append these tags to your prompt when you want to summon the funnels:

Plaintext

avalanche (weapon), floating weapon, funnel (weapon),

⚙️ Weight Recommendations & Testing Notes

• Recommended Strength: 0.4 ~ 0.6

• Clothing Swap Stress Test: I ran a stress test using the Prompt Generator 12 Columns with 50 random outfit generations. When paired with anima-turbo-lora-v0.2 (at strength 0.9), a LoRA strength of 0.4 yields the best results for custom clothing swaps (though not guaranteed every single time).

• ⚠️ Concept Bleed Warning: If the LoRA strength is lowered to 0.3 or below, the character features will start to bleed and mix with the homonymous character "Firefly (Star Rail)" baked into the Anima Base V1 foundation model. Therefore, it is highly recommended to keep the strength above 0.4 to maintain character accuracy.

The above was translated by Gemini.

下面中文介紹

Firefly (StellaSora) LoRA

這是一個專為 Firefly (StellaSora) 訓練的角色 LoRA。支援多套原創服裝與武器配件，並經過完整的換裝壓力測試。

🔑 核心觸發關鍵字 (Trigger Words)

• 角色基礎：firefly_(stellasora)

• 日常衣服 (Explorer Costume)：stella_explorer_(costume), blue_scarf, small pink camera

• 夏洛克企鵝裝 (Sherlock Penguin)：sherlock_(penguin_costume), penguin_pajamas

• 白色早期衣服 (Early Nightgown)：sora_pajamas_(nightgown), white_nightshirt

• 雪崩武器 / 浮游砲 (Avalanche Weapon)：avalanche_(weapon), floating_weapon, funnel_(weapon)

💡 建議提示詞格式 (Suggested Prompts)

建議直接複製以下基礎格式進行微調：

日常衣服 (Daily Explorer Outfit)

Plaintext

1girl, firefly (stellasora), stella explorer (costume), blue scarf, short hair, light brown hair, red eyes, pink small camera on belt,

夏洛克企鵝裝 (Sherlock Penguin Suit)

Plaintext

1girl, firefly (stellasora), sherlock (penguin costume), penguin pajamas, ahoge, blue scarf, red eyes, bare legs,

白色早期衣服 (Early White Nightgown)

Plaintext

1girl, firefly_(stellasora), sora_pajamas_(nightgown), white_nightshirt, oversized_shirt, short hair, light brown hair, wide_sleeves, long_sleeves, barefoot, white choker,

裝備武器 (Equipping Weapon) 需要生成武器時，請在上述格式後方疊加以下標籤：

Plaintext

avalanche (weapon), floating weapon, funnel (weapon),

⚙️ 模型權重與測試建議 (Usage & Testing Notes)

• 推薦權重 (Recommended Strength)：0.4 ~ 0.6

• 換裝與泛用性測試：我使用了 Prompt Generator 12 Columns 隨機生成 50 種服裝進行測試。在同時掛載 anima-turbo-lora-v0.２ (Strength: 0.9) 的環境下，本 LoRA 的權重設定在 0.4 時，最容易成功進行自定義換衣（儘管無法保證 100% 成功）。

• ⚠️ 特徵污染警告 (Concept Bleed)：請注意，如果將本 LoRA 的權重降至 0.3 或更低，潛空間的概念會開始與 Anima Base V1 基礎模型中同名的「流螢 Firefly 崩壞：星穹鐵道」特徵產生混合。為了確保角色特徵的正確性，強烈建議權重保持在 0.4 以上。