This model focuses on "fast character portrait creation + cinematic lighting clarity", achieving a natural balance between 2D illustrations and realistic rendering. It can output character illustrations, game CG, and virtual avatar materials that meet commercial-grade visual standards.

Style: Compatible with Japanese-style illustrations, ACG character portraits, and 3D rendering, supporting diverse visual expressions from golden-hued 2D styles to semi-realistic aesthetics.

Detail Precision: Features extremely high restoration of facial features, hair texture, wrinkles, and skin texture. It can generate light and shadow scenes with rich layers and strong atmospheric sense.

Application Scenarios

Game Industry: Character portraits, card illustrations, promotional posters, NPC image design, skin and fashion original art.

Cultural and Creative Content Creation: Fan art, high-definition wallpapers, peripheral product (badges/standees) materials, VTuber design drafts, novel covers, comic storyboards.

Commercial Marketing: Brand co-branded meta-posters, product packaging illustrations, social media visual materials, live broadcast virtual backgrounds.

Virtual Idol Ecosystem: Virtual idol character design sheets, live dynamic backgrounds, emoji materials, derivative short video visual assets.

Image Generation Parameter Reference

Basic Generation Parameters

| Parameter Item | Recommended Value | Description |

|----------------|-------------------|-------------|

| Sampler | DPM++ 2M Karras / Euler a | Balances speed and detail performance |

| Sampling Steps | 20–30 steps (25 steps recommended) | More steps yield richer details; 25 steps balance efficiency and quality |

| Resolution | 1024×1536 / 1536×1024 | Supports vertical portrait/illustration ratios and 2K output |

| Aspect Ratio | 2:3 (character portraits) / 1:1 (avatars/cards) | Matches the visual ratio of reference images |

| Prompt Weight | Positive 1.0–1.2 / Negative -1.0~-1.5 | Preserves high-quality features and avoids low-quality elements |

Positive Prompt Template (for unknown reference image styles)

Plain Text

(masterpiece, top quality, ultra-detailed), (anime CG, realistic painting style), 1 girl, (rich facial details, delicate skin texture, glossy hair), (exquisite clothing design, delicate fabric texture), dramatic lighting effects, cinematic composition, (dynamic pose, vivid expression), deep background, (soft shadows, high contrast), professional illustration

* Character features can be replaced as needed (e.g., "1boy", "elf ears", "mecha armor", etc.)

Negative Prompt Template (to avoid low-quality results)

Plain Text

low resolution, bad anatomy, bad hands, text, error, missing fingers, extra fingers, fewer digits, cropped, worst quality, low quality, normal quality, JPEG artifacts, signature, watermark, username, blurry, ugly, mutated, deformed, disfigured, poorly drawn face, extra limbs, floating limbs, disconnected limbs, malformed hands, long neck, duplicated, mutilated, diseased, out of frame

Advanced Optimization Parameters

• Archaeological Restoration (Hire. Repair): Enable, magnification 1.5–2.0, redraw amplitude 0.3–0.5 (improves detail sharpness)

• Face Restoration: Enable CodeFormer, strength 0.6–0.8 (optimizes facial precision)

• Enhancement: Latent Upscale + Sharpening Filter (improves image clarity)

• Seed: Fixed seed for reproducible results; random seed for diversified content

DQ Semi-Realistic Asian CG V1 - 1.0