Melon-Bread/anima-pc98-gal-style-v0.1

PC-98 Gal

Prompt

pc98gal_style, 1990s anime aesthetic, retro PC-98 game art, limited color palette, flat cel shading, thin linework, detailed retro eyes, dithering, muted colors, nostalgic atmosphere, beautiful detailed background, 1girl, silver hair, green eyes, ornate plate armor, standing in ancient ruins covered in ivy, dramatic stormy sky, holding glowing sword, full body, heroic pose

Negative Prompt

3d, realistic, photorealistic, modern anime style, thick outlines, gradient mesh shading, oversaturated, blurry, deformed, extra digits, watermark, text, signature

Model description

PC-98 Galgame Style (Anima)

Trained words: `pc98gal_style`

Base model: Anima v1.0

---

Description

A LoRA trained to reproduce the aesthetic of NEC PC-98 galgame artwork from the early-to-mid 1990s. Captures the signature look of era-defining eroge/visual novel studios — flat cel shading, thin linework, muted limited color palettes, detailed retro anime eyes, and atmospheric painted backgrounds.

Dataset consists of 35 images sourced from 14 authentic PC-98 titles including Words Worth, Nocturnal Illusion, Doukyusei 2, Yuno, Viper GTS, and others, covering a range of character designs and settings.

Recommended Settings

• LoRA strength: 0.7–1.0
• Steps: 25–30
• CFG: 4.0–5.0
• Sampler: euler, scheduler: simple
• Resolution: 832×1216 (portrait), 1216×832 (landscape)

Usage

Place `pc98gal_style` at the start of your prompt, followed by your subject description.

Usage

Place pc98gal_style at the start of your prompt, followed by your subject description. Here are some other recommended prompts to add

Positive prompt:

```

pc98gal_style, 1990s anime aesthetic,, limited color palette, flat cel shading, thin linework, detailed retro eyes, dithering, muted colors,

```

Negative prompt:

```

3d, realistic, photorealistic, modern anime style, thick outlines, gradient mesh shading, oversaturated, blurry, deformed, extra digits, watermark, text, signature

```

Notes/Ramblings

This is my very first time every making a lora or really doing any kind of training so expect noobie mistakes. I do not think I am done with this as I want to work on it some more to get better results. Though figured I should share my early results now to keep me motivated to keep going.

Network Specs

• LoRA rank/dim: 32, alpha: 32
• Trained on both DiT and text encoder
• 10 epochs / 1050 steps / final loss 0.128
• Training time: ~55 min on Radeon 8060S (gfx1151)

Trigger words

You should use pc98gal_style to trigger the image generation.

Download model

Download them in the Files & versions tab.