CivArchive
    SHIFUKU Gold Leaf v2 — Kiribaku / Sunago / Noge | Physical Texture LoRA [SDXL] - v2.0 — Kiribaku / Sunago
    Preview 126180860
    Preview 126180868
    Preview 126180863
    Preview 126180862

    What is this?

    A LoRA that captures the physical texture of real Japanese gold leaf (金箔 / kinpaku) — not the color, not the "gold style," but the actual material surface.

    v2 update: Now trained on expanded dataset including 3 specialized gold leaf application techniques used in traditional Japanese painting:

    TechniqueJapaneseWhat it looks like Kiribaku (切箔)切箔Small squares of gold leaf scattered across the surface — visible edges, overlapping layers Sunago (砂子)砂子Gold dust sprinkled as fine particles — grainy, atmospheric, soft shimmer Noge (野毛)野毛Thin strips of gold leaf — linear, directional, like golden rain

    These aren't filters or effects. They're physical application methods developed over centuries by Japanese artists. Each creates a completely different surface texture from the same material.

    What's New in v2

    v1v2 Dataset50 reference imagesExpanded with Met Museum collection references TechniquesGeneral gold leaf surfaceKiribaku, sunago, noge + general Grain detailGoodSignificantly improved — individual leaf edges visible Application varietySingle surface typeMultiple traditional techniques Light interactionBasic scatterEnhanced — different behavior per technique

    v1 captured "gold leaf as a surface." v2 captures "how gold leaf was applied" — and the texture changes completely depending on the technique.

    The Problem This Solves

    90% of AI-generated gold backgrounds are flat. They lack:

    • The visible grid lines where square leaf sheets meet (kiribaku)

    • The atmospheric depth of scattered gold dust (sunago)

    • The directional energy of gold leaf strips (noge)

    • The oxidation patterns that shift with age

    • The "haku-ashi" (箔足) — the raw edge where leaf was torn by hand

    v1 taught the model that gold leaf has texture. v2 teaches it which kind of texture.

    How It Was Trained

    Reference images sourced from museum collections (The Metropolitan Museum of Art Open Access) and original photography.

    3-Distance Method:

    DistanceWhat it captures Extreme close-upIndividual gold particles, leaf edge sharpness, oxidation micro-pattern Mid-rangeApplication technique visible (kiribaku grid vs sunago scatter vs noge lines) Full viewGold leaf as background material in traditional painting context

    Trained with Kohya_ss on SDXL.

    Sample Prompts

    Kiribaku (scattered squares) — bird on gold screen:

    sparrow perching on bamboo branch, gold leaf background, kiribaku scattered gold squares, traditional japanese screen painting style, <lora:SHIFUKU_GOLDLEAF_v2:0.8>

    Negative: smooth, flat gold, digital gradient, plastic
    Settings: Steps 35 / CFG 7 / Sampler DPM++ 2M Karras

    Sunago (gold dust) — atmospheric cherry blossoms:

    cherry blossom branches, gold dust background, sunago technique, soft shimmer, traditional nihonga painting, atmospheric depth, <lora:SHIFUKU_GOLDLEAF_v2:0.7>

    Negative: flat, uniform, modern, digital
    Settings: Steps 40 / CFG 7.5 / Sampler Euler a

    Bold gold leaf — painterly bird:

    bird on branch, gold leaf background with visible texture, thick impasto brushwork, traditional japanese art meets western oil painting, <lora:SHIFUKU_GOLDLEAF_v2:0.85>

    Negative: smooth, plastic, flat color
    Settings: Steps 38 / CFG 7 / Sampler DPM++ 2M Karras

    Who Made This

    TextureLoRALab — I studied Japanese painting (nihonga) in art school, earned a Master's degree in Museum Studies (with Merit) from a UK university, and now I combine traditional material knowledge with AI training techniques.

    The difference between kiribaku, sunago, and noge isn't something you learn from Google. It's something you learn from looking at actual screens and scrolls under museum lighting. That material knowledge shaped this dataset.

    • Base model: DreamShaper XL alpha2 (works best)

    • Weight: 0.6–0.8 for subtle texture, 0.9–1.0 for full effect

    • v2 tip: Mention specific techniques in prompt (kiribaku, sunago, noge) for targeted results

    • Works with: Portraits, birds-and-flowers, landscapes, still life, abstract

    Trigger Words

    gold leaf texture, kinpaku, kiribaku, sunago, noge

    Commercial Use

    ✅ Free for commercial use. No credit required.

    More Textures in the SHIFUKU Series

    TextureWhat it capturesStatus Gold Leaf v1 (金箔)Basic gold leaf surface texturelegacy — use v2 Gold Leaf v2 (金箔)Kiribaku, sunago, noge techniquesthis model Kintsugi (金継ぎ)Sharp gold repair lines on matte ceramicavailable Hamon Steel (刃文)Temper line, blade polish, lacquer scabbardavailable (beta)

    Follow TextureLoRALab on CivitAI to get notified.


    "Style LoRAs teach AI how things look. Matière LoRAs teach AI how things feel."

    Description

    v2.0 — Kiribaku / Sunago / Noge

    Major update: expanded dataset with 3 specialized gold leaf application techniques from traditional Japanese painting.

    • New techniques: Kiribaku (切箔/scattered squares), Sunago (砂子/gold dust), Noge (野毛/thin strips)

    • Dataset expanded with Met Museum Open Access Collection references

    • Significantly improved grain detail — individual leaf edges now visible

    • Enhanced light interaction — different behavior per technique

    • Multiple traditional application methods instead of single surface type

    v1 captured "gold leaf as a surface." v2 captures "how gold leaf was applied."

    LORA
    SDXL 1.0

    Details

    Downloads
    9
    Platform
    CivitAI
    Platform Status
    Available
    Created
    4/2/2026
    Updated
    4/2/2026
    Deleted
    -

    Files

    SHIFUKU-KINPAKU-GOLDLEAF.safetensors

    Mirrors