MOSS-Image-Base (5B) - BAStyle LoRA (v2.0)

A BF16 Blue Archive style crisp cel-shaded illustration LoRA (v2.0 Facial & Visual Fidelity Edition) trained on OpenMOSS / MOSS-Image-Base (5B).

OpenMOSS / MOSS-Image-Base (5B) 的 BF16 蔚蓝档案风格高清赛璐璐画风 LoRA（v2.0 五官与画质精炼版）。

Note: All showcase images were generated raw with BAStyle v2 alone on the official 8-bit (moss_image_base_int8_convrot) base model. Running on BF16 shows minor detail differences, and lower-bit quantizations (such as INT4) or Turbo variants have not been tested.

说明：所有展示样张均在官方 8-bit (moss_image_base_int8_convrot) 底模上单挂 BAStyle v2 原生直出实现。BF16 下会有细微细节差异，暂未在更低比特量化（如 INT4）或 Turbo 版本上实验。

What's New in v2.0 / v2.0 更新说明

• Improved Standalone Facial Proportions (单跑人物五官表现改善): Addresses v1.0's issue where standalone original characters (including short or Chinese prompts) could retain older anime facial proportions and overly glossy hair highlights. By refining bangs-cut, gaze, blush, and expression annotations across all 58 training illustrations, v2.0 significantly improves standalone facial geometry and matte cel-shaded hair rendering.
  针对 v1.0 单跑路人角色（尤其是中文或简短提示词）偶发早期二次元脸型与油亮高光发质的问题，通过对 58 张高清原画的刘海切法、眼神、腮红与神态精细化重标注，显著改善了单跑角色的五官比例与哑光赛璐璐发质表现。

• Higher Contrast & Sharpness (v2.0 vs v1.0 Tip): v2.0 delivers sharper cel-shading edges and stronger color contrast than v1.0. If you prefer v1.0's softer, lower-contrast airy pastel shadow floor (or when stacking heavy character LoRAs), use v1.0 or lower v2.0 strength to 0.80 - 0.85.
  v2.0 线稿更利落，明暗对比与色彩饱和度比 v1.0 更鲜明；若更偏爱 v1.0 柔和清淡的低对比度调性，可继续选用 v1.0 或将 v2.0 权重调至 0.80 - 0.85。

Overview / 简介

• Style & Scene Coverage: Covers both close-up Memorial Lobby portraits and even-lit story cutscene CGs (multi-character group scenes, lone character overlooking a cityscape, campus architecture, and daytime/nighttime skies).
  除近景纪念大厅插画外，同样支持清爽平光的剧情过场 CG（多人日常同框、单人背影远眺）以及学院建筑、昼夜天空等背景空景。

• Halos: Not baked into the trigger word. Describe halo color and shape when desired; omit halo terms for halo-free characters.
  光环按需控制：需要光环时描述颜色与形状；不写光环词汇时默认不出光环（男性、奇幻服饰等均零泄漏）。

• Pure Empty Backgrounds / Skies: When generating scenes with no people in the frame, include 2D anime background illustration in the opening clause (instead of photographic shot) so the base model does not fall back to realistic photography. Add with faint concentric rings to summon the signature ringed sky.
  画面无人物的纯建筑或天空空景建议在开头加入 2D anime background illustration 以避开底模写实摄影先验；描写天空时加入 with faint concentric rings 可生成原作标志性的同心圆环天空。

• Frozen Text Encoder: Trained on the DiT only (Rank 32 / Alpha 16, BF16) with Qwen3.5-9B frozen, preserving native natural-language prompt adherence and in-image text rendering.
  冻结 Qwen3.5-9B 仅训练 DiT，保留底模原生的中英文自然语言理解与文字生成能力。

ComfyUI Settings / 推理参数

A dependency-free native ComfyUI workflow is attached as a Workflow .zip file (and embedded directly inside the sample PNGs).

附带无第三方插件依赖的原生 ComfyUI 文生图工作流压缩包（样张原图内同样嵌有完整工作流元数据，下载原图拖入 ComfyUI 即可直接加载）。

• Base DiT (UNETLoader): moss_image_base_bf16.safetensors or moss_image_base_int8_convrot.safetensors

• Text Encoder (CLIPLoader): qwen3.5_9b_int8_convrot.safetensors (or bf16), Type: moss_image

• VAE (VAELoader): flux2-vae.safetensors

• Latent (EmptyFlux2LatentImage): Multiples of 16: 1344x896, 1536x1024, 896x1344, 1920x1088 (recommended >= 960x640)

• Trigger Word: bastyle, at the start of the prompt

• LoRA Strength: 0.90 - 0.95 standalone (0.80 - 0.90 with character LoRAs)

• Flow Shift (ModelSamplingAuraFlow): 5.0 (6.0 above 3 MP)

• Skip-Layer (SkipLayerGuidanceDiTSimple): double_layers = "", single_layers = "22", start = 0.0, end = 1.0

• Sampler / Scheduler: euler + simple, 28 - 32 steps, CFG = 3.0

• Negative Prompt: Empty (""), or halo, glowing halo above head if strictly suppressing halos

Prompting Guide & Examples / 提示词写法与完整示例

Do not use Danbooru-style tag lists (such as masterpiece, best quality, 1girl). Write complete natural-language sentences starting with bastyle,.

无需堆砌 masterpiece, best quality, 1girl 等传统词条，请在 bastyle, 后直接使用完整的自然语言长句描述画面。

Prompt Templates / 万能句式结构

• Single Character / Memorial Lobby (单人 / 纪念大厅)
  bastyle, [Camera angle & shot size] of [hair/eye color, expression, action, optional halo color & shape], wearing [outfit details] in [environment], lit by [light type & direction].

• Multi-Character Cutscene CG (多人剧情过场 CG)
  bastyle, [Wide/medium-wide shot] of [N] students in [scene], lit by clear even daylight from [direction]. Left to right: [Character A details & halo]; [Character B details & halo]; [Character C details & halo].

• Empty Architecture / Sky Background (纯无人建筑 / 天空空景)
  bastyle, A [camera angle] wide 2D anime background illustration of [architecture or sky details + optional "with faint concentric rings"], [lighting], no people in the frame.

Verified Sample Prompts / 实机验证示例提示词

1. Memorial Lobby - With Halo (纪念大厅 · 带光环)

bastyle, An eye-level medium shot of a quiet girl with long dark-violet hair, warm amber-red eyes, and a thin glowing pale-pink geometric halo above her head sitting by a tall sunlit library window, holding a closed hardcover book against her chest and looking calmly toward the viewer. She wears a cream knit cardigan over a white collared shirt and navy pleated skirt, with soft afternoon sunlight streaming through the window from the left.

2. Memorial Lobby - Without Halo (纪念大厅 · 无光环)

bastyle, An eye-level medium close-up shot of a gentle girl with flowing silver-white hair, luminous sapphire-blue eyes, and a small navy hair ribbon leaning over a sunlit seaside boardwalk railing at golden hour sunset, smiling softly back toward the viewer as the warm sea breeze lifts her hair and red neck ribbon. She wears a crisp cream cardigan over a white school shirt, bathed in translucent amber-gold sunset rim light from the upper left against sparkling ocean waves and pastel orange-and-violet twilight clouds.

3. In-Image Text Rendering (画面内英文写字)

bastyle, A slight low-angle medium close-up shot of an energetic blonde wolf-eared girl with bright emerald-green eyes, a playful fang smile, and a glowing golden star-ring halo floating above her head standing on a breezy school rooftop under a crisp azure sky with fluffy cumulus clouds, holding up a small wooden cafe chalkboard cleanly written in neat white chalk text "After-School Sweets!" with a cute doodle star beside it, wearing an open pastel-mint hoodie over a white blouse with bright afternoon sunlight from the upper right.

4. Story Cutscene CG - Lone Overlook (单人背影远眺城市过场)

bastyle, A wide eye-level shot from behind a lone girl with long silver-blue hair in a navy school cardigan standing at a metal observation deck railing in the right foreground, looking out across a sprawling coastal city, harbor bridges, and distant mountain ridges under a calm late-afternoon sky, lit by gentle natural daylight from the upper left.

5. Story Cutscene CG - Multi-Character Group (平光多人街头过场)

bastyle, An eye-level medium-wide group shot of three students walking abreast across a clean city street crosswalk before low urban buildings and green street trees under a blue sky, lit by clear even daytime sunlight from the upper left. Left to right walk a short-haired blonde student with green eyes and a yellow ring halo waving one hand; a dark-haired student with round glasses and a cyan geometric halo carrying a paper shopping bag in a navy blazer; and a pink-ponytail student with amber eyes and a pale-pink halo holding a canned drink in a beige cardigan and grey plaid skirt.

6. World Background - Campus Architecture (学院建筑无人空景)

bastyle, A symmetrical eye-level wide 2D anime background illustration across a clean white stone plaza and classical colonnade courtyard leading toward a grand modern-gothic academy clock tower and glass-walled campus buildings under a clear blue sky with faint concentric rings, crisp cel-shaded architectural linework and clean cast shadows across pale stone pavement, no people in the frame.

7. World Background - Daytime Ringed Sky (白天同心圆环天空)

bastyle, A low-angle wide 2D anime background illustration looking up past a painted school rooftop railing and wire fence into a vibrant clear cyan-blue daytime anime sky filled with painterly white cumulus clouds and giant faint concentric halo rings across the upper sky, bright clean anime daylight, no people in the frame.

8. World Background - Nighttime Ringed Sky (黑夜同心圆环星空)

bastyle, A wide nighttime low-angle 2D anime background illustration from a quiet hillside road and white guardrail looking up at a deep-indigo starry anime night sky with faint concentric rings, glowing stars, and a crisp crescent moon above stylized dark mountain silhouettes and warm distant town lights in the valley below, cool moonlight, no people in the frame.

9. Direct Chinese Prompt Example (纯中文自然语言直出)

bastyle, 平视中近景镜头，清晨洒满阳光的学院林荫道上，一位留着深紫色长发、紫晶色眼眸、戴着圆框眼镜的温柔少女抱着一摞精装书回头微笑，身着米色针织开衫、白色衬衫与深蓝色百褶裙，微风轻轻吹起她的发丝，左上角通透的晨光在发梢与肩头勾勒出明亮的高光。

Known Limitations & Roadmap / 已知问题与后续计划

• Higher Contrast & Saturation in v2.0 (v2.0 对比度偏高说明): Because the v2.0 captioning pass included fine-grained color-shade and inner-shadow modifiers, part of the soft shadow lift was decoupled into text, resulting in noticeably higher contrast and saturation than v1.0. Since the overall visual clarity and facial rendering still look clean and appealing on their own, v2.0 is released for users who enjoy a crisper, higher-contrast look.
  由于 v2.0 打标时细化了较多色彩明度与暗部阴影修饰词，使得出图整体对比度与饱和度比 v1.0 偏高。不过实际成图画质与五官表现依然干净耐看，因此还是决定发出来供喜欢鲜明高对比画风的朋友使用。

• v3.0 Contrast Recalibration Plan (v3.0 后续修正计划): v3.0 will strip excess color-value/shading modifiers from the training captions to restore v1.0's signature low-contrast, airy pastel look while keeping v2.0's improved facial structure.
  下一版 v3.0 将针对性调整对比度偏高的问题，在保留 v2.0 五官结构优化的同时恢复 v1.0 清透柔和的低对比度调性。

License & Disclaimer / 协议与免责声明

Released under CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0). This LoRA is an unofficial fan-made style study intended strictly for non-commercial personal learning, research, and fan creation. Any commercial use, paid generation service, or paid redistribution is prohibited.

本模型权重采用 CC BY-NC 4.0（署名-非商业性使用 4.0） 协议发布，仅限个人非商业性质的技术研究、学习交流与同人二创使用，严禁用于任何商业用途、付费代出图或二次售卖。