CivArchive
    YALL - YetAnoterLatexLora - v0.01H3
    NSFW

    YALL - YetAnotherLatexLora (v0.04)

    a latex fetish LoRA for Krea2

    Trigger: yall

    The trigger on its own just means "latex fetish". Every material and garment is bound to the words, so name what you want: latex, the finish (high-gloss), a colour, and the pieces. Everything below only shows up when you ask for it.

    v0.04 is a big step up from v0.03. The dataset grew from 443 hand-picked stills to 10,487 captioned images, all with ~100-word natural-language captions, and it is trained at rank/alpha 128 on Krea2 Raw. The main prompting change is that you should write fresh prompts instead of copying the caption wording, more on that below.


    ⚙️ Settings

    • Krea2 Turbo: sampler euler, scheduler simple, 8 steps, cfg 1.0 (fixed, do not change).

    • Resolution: example renders are ~1.5MP: 16:9 (1664x928) / 9:16 (928x1664); 1024x1024 and other 1MP ratios also work.

    • LoRA strength: 0.7 to 0.9, the example images use 0.9. Higher gives more gear detail, around 0.7 looks cleaner and more natural.

    • Prompt: plain natural-language prose with the trigger at the front, 120-150 words works best.

    Changed since v0.03: don't copy the training caption wording. Prompts that echo the captions tend to come out soft and degraded.


    🧠 What it knows

    Strongest to weakest, by how often it appears in the training captions.

    Core (always reliable, still name them in the prompt)

    • finish: high-gloss (99.8% of captions), with sharp specular highlights and reflections of the room

    • colour: black (by far the strongest)

    • garment: latex catsuit

    Strong (reliable when you ask for it)

    • latex gloves

    • face covering: a latex hood or mask (both are about equally common now), either faceless (covering the entire head and face) or open-face (leaving only the eyes, or the eyes and lips, exposed)

    • the material word "rubber" as well as "latex"

    • silicone reptile-scale texture

    • restraint hardware: collar

    • poses: standing

    • two people in one picture (give them different outfits and poses, "matching" outfits in mirrored poses come out as clone twins)

    Solid (works, a little less reliable than Strong)

    • garments: boots, latex bodysuit, heels / platform, corset

    • colour: red, then pink (pink works as a suit colour now, it was weak in v0.03)

    • translucent latex

    • restraint hardware: straps, buckles, zippers

    • silicone suits

    • breathing gear: gas mask

    • finish: wet-look

    • colour: white, yellow (often translucent)

    • body on show: breasts / cleavage cues

    Occasional (works, less consistent)

    • garments: dress, skirt, stockings, leggings

    • restraint: harness, chains, bound poses

    • colour: blue, green, purple

    • creature hoods: pup / puppy, crown / spikes / horns

    • gags (ball gag strongest)

    • poses: kneeling

    • settings: bedroom, industrial / concrete, studio, bathroom / shower, mirror

    • toys: dildo

    Rare (mostly the base model's idea of it, name at your own risk)

    • mirror-like finish, vibrator, massager / wand, ring gag, kigurumi masks (avoid the word "doll")

    • open-crotch / crotchless, bare skin / nudity

    • inflatable / inflated latex: ball hood, ball mitts

    • bunny hood, tail, cape, maid, nurse

    • heavy / thick rubber phrasing

    Not in the v0.04 data (you get the base model's idea of it)

    • hobble, sleepsack

    (Some things v0.03 listed as "not in the data" now show up a little, like vacuum bed, rope, cage, straitjacket, armbinder, a man and walking, but they are thin and unreliable.)


    📝 Usage notes

    • Write fresh, natural photographic prompts about what you want to see; do not reuse the training-caption phrasing (it renders soft). Keep yall, at the front.

    • Add an age and a build yourself: a slim, toned young woman in her twenties. The captions never state an age, and a bare "a woman" can drift older. "curvy" pushes the body heavier.

    • Say "high-gloss": it is in 10,468 of the 10,487 captions. "glossy" and "shiny" are much weaker.

    • Faceless look: a latex hood that covers her entire head and face. Naming eye or mouth openings brings the face back. Eyes-and-mouth hoods especially tend to render as an open human face, so use fully sealed wording if you want a sealed look.

    • "doll" tends to give a literal doll or mannequin.

    • Two people: give them clearly different outfits and poses ("the woman on the left ... the woman on the right ..."); matching outfits in mirrored poses come out as clone twins. Don't use the word "second", it tends to add an extra person.

    • Explicit acts and rare gear (mitts, armbinder, vacuum bed, inflatables, toys) are thin in the data, so you mostly get the base model's take on them.

    • Text quirk: "yall" can occasionally print as literal lettering (plain walls, signs, collar tags, pup-mask markings). cfg is fixed at 1.0, so a negative prompt can't remove it. Change the seed or drop the words that invite lettering.


    🖼️ Examples

    Red hood editorial

    yall, A full-body editorial frame, shot slightly below eye level on a wide prime lens, of a slim young woman in her twenties poured into a high-gloss red latex catsuit. A seamless red latex hood of the same rubber seals her whole head and face, a smooth featureless shell with no openings. Long black latex gloves climb past her elbows, and a black latex corset laced with silver hooks draws her waist in tight. She stands square to the camera with one gloved hand resting at her hip. The rubber stretches drum-tight over every curve, throwing hard specular glints and mirroring the pale room back across her shoulders and thighs. The backdrop is a bare lilac plaster wall over a grey stone floor. A broad softbox from the left wraps her in cool, even light, the palette all red, black and dusk-violet.

    Forest pony hood

    yall, A low-angle full-body shot on a 35mm lens of a young woman in her twenties standing among tall pine trunks in a quiet forest. She wears a high-gloss two-tone latex catsuit, black across the chest and arms with olive-green panels running down the torso and hips. A black latex pony hood with upright pointed ears seals her whole head, its integrated gas-mask facepiece carrying two round glass eyepieces and a short filter below. Both gloved hands lift to rest flat on either side of that mask. The latex pulls tight over her frame, catching sharp specular highlights and holding faint green reflections of the trees. Shafts of bright diffused daylight fall between the trunks onto the forest floor, the palette built from forest green, warm brown and black.

    Pink corset studio portrait

    yall, A medium studio portrait on a short telephoto lens of a young woman in her twenties, framed from the thighs up against a seamless charcoal paper backdrop. She wears a wet-look black latex catsuit with a long front zipper, the finish so high-gloss it reads almost liquid under the lights. Over it a pink latex underbust corset with a row of small buckles cinches her waist, and a slim black collar sits at her throat. A black latex hood of the same rubber seals her whole head and face in one smooth, featureless piece. She tilts her chin down with both gloved hands clasped loosely in front of her. Twin softboxes rake the rubber from either side so the highlights run in bright vertical streaks down her arms and ribs, the palette kept to black, soft pink and grey.


    I also created a discord server with a bot with my model(s), where you can have fun with my lora's for free, and chat with other users, or even request some loras? who knows 😉 feel free to join 😊 https://discord.gg/ngE2g3hvsE

    Description

    Experimental version of Minimax H3 model, could be prompted with more or less same verbiage as YALL main lora , this is an early version of dataset and lora so if you do not prompt against music it will be likely added because i forgot to trim audio in clips 😁

    FAQ

    LORA
    MiniMax H3

    Details

    Downloads
    228
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/11/2026
    Updated
    10/13/2026
    Deleted
    -

    Files

    yall_h3_video_v1.safetensors

    Mirrors

    CivitAI (1 mirrors)