FasciumERNIE

FASCIUMFASCIUM
36
501
base model
FINEDNEWTURBO_WOMANIZEDTURBO_EXTENDEDturbo
model-image-0
model-image-1
model-image-2
model-image-3
model-image-4
model-image-5
model-image-6
model-image-7
model-image-8
model-image-9
model-image-10
model-image-11
model-image-12
model-image-13
model-image-14
model-image-15
model-image-16
model-image-17
model-image-18
model-image-19

μ„ΈλΆ€ 정보

파일 λ‹€μš΄λ‘œλ“œ (1)

λͺ¨λΈ μ„€λͺ…

🎞️ FasciumERNIE Turbo

Cinematic Historical Portraits & Analog Film Aesthetics

A highly specialized DiT-based checkpoint engineered for cinematic historical portraiture, vintage photography, and atmospheric storytelling. Merged from ERNIE-Image base with carefully weighted European facial feature LoRAs and analog/filmic styling adapters. Delivers museum-grade realism, rich textile rendering, and authentic period aesthetics in just 8–10 steps.

βš™οΈ Recommended Settings

Parameter

Value

Sampler

Euler

Schedule Type

Simple

Steps

8–10

CFG Scale

1.0

Shift

3.0

Resolution

1024Γ—1536 (or 896Γ—1152)

Diffusion Low Bits

float8-e5m2 (highly recommended)

VAE

Auto / flux2-vae

Text Encoder

ministral-3-3b (or default)

πŸ”‘ Golden Rule: Keep CFG at 1.0 and use Euler + Shift 3. Higher CFG or different samplers break the Turbo balance and cause oversaturation/artifacts.

🎨 Style & Strengths

βœ… Analog/Cinematic Aesthetic: Built-in film grain, natural color grading, and optical imperfections (vignette, soft focus, chromatic aberration).
βœ… Historical & Period Accuracy: Excels at Viking, Byzantine, Celtic, Slavic, Art Deco, Rococo, Victorian, and Medieval eras.
βœ… Textile & Material Mastery: Velvet, silk, lace, fur, aged leather, and metallic embroidery render with tactile realism.
βœ… Emotional Portraiture: Captures subtle expressions, atmospheric lighting (candlelight, fog, rim light), and narrative depth.
βœ… Lightning Fast: Turbo-optimized for 8-step generation without sacrificing artistic quality.

⚠️ Limitations & Pro Tips

πŸ”Ή Hands & Small Objects: At CFG 1.0, complex hand poses or tiny props may occasionally glitch. Fix: Use Euler + Shift 3, prompt wearing gloves / hands clasped / resting on surface, or increase steps to 10.
πŸ”Ή Not for Macro/Mechanics: Optimized for portraits & scenes. Avoid prompts focused on intricate machinery, insects, or abstract physics.
πŸ”Ή Style Control: The model thrives on descriptive prompts (cinematic film still, Baroque lighting, film grain). Avoid overloading with negative tags.
πŸ”Ή Negative Prompt (Optional):
cartoon, anime, 3d render, plastic, doll, bright, happy, lowres, bad anatomy, text, watermark

🏷️ Trigger Words

fascium_style, cinematic portrait, analog photography, historical realism, film grain, baroque lighting

πŸ“œ Technical Details

  • Base Model: ERNIE-Image (DiT Architecture)

  • Merge Method: Weighted sum (alpha=0.3) + 70/30 base blend

  • Training Focus: European facial structures, period costumes, vintage film emulation

  • File Size: ~15 GB (FP16) / ~6 GB (FP8 optimized)

  • Architecture: Diffusion Transformer (S3-DiT variant)

  • Tested On: Forge Neo / WebUI 1.10+

πŸ™ Tested extensively across 10+ historical eras. If you create something amazing, drop a generation in the comments!
πŸ”₯ For best results, stick to the recommended settings. Deviating may break the Turbo balance.


이 λͺ¨λΈλ‘œ λ§Œλ“  이미지