CivArchive
    Masked Fontaine series 美聖女仮面フォンテーヌ (Realistic, semi-realistic JAV rendition) - Aurora v1
    NSFW
    Preview 12350793
    Preview 12350773
    Preview 12350780
    Preview 12350784
    Preview 12350789
    Preview 12350786
    Preview 12350770
    Preview 12350782
    Preview 12350774
    Preview 12350785
    Preview 12350777
    Preview 12350778
    Preview 12350792
    Preview 12350779
    Preview 12350787
    Preview 12350791
    Preview 12350772
    Preview 12350788
    Preview 12350771
    Preview 12350790

    This is a costume/character LoRA to reproduce the likeness of the popular Fontaine heroine (美聖女仮面フォンテーヌ) costume based on the JAV renditions. Now you can complete your heroine fetish dreams.

    Been sometime since I do something focusing on JAV heroines, so here it is.

    Aurora Version 1

    -Introducing Fontaine Aurora (the pink Fontaine).

    Fontaine Aurora is another heroine character within the Fontaine JAV GIGA-franchise. Story-wise, Aurora's backstory is slightly separate from the original Fontaine, and its more like spin-offs of small stories, with slightly lesser consistency. Then again, only the Spandexe-franchise respects the timeline of the plot, the Fontaine series treat most stories like a standalone episode.

    Typically, Fontaine Aurora was played by slightly young and cute looking actresses (typical mahou shoujo vibes), but occasionally, the director do get mature looking actresses to play them. In my training set, I used both - but I train it with great level of flexibility that you should be able to create any both versions without any issues.

    I have also tokenise each component of her attire as detailed as possible, so you should be able to mix-and-match her costumes (for NSFW variants etc.)

    Proper:

    1girl, f0nta1ne_4urora, fontaine_mask, fontaine_hat, capelet, white footwear, knee boots, cleavage cutout, cleavage, white gloves, white dress, belt, bowtie, <lora:Masked_Fontaine_aurora_v1_epoch_12:0.8>

    weapon-holding, magic:

    magical wand, holding wand, sword, holding weapon, special effect, magic

    Tips: use cleavage cutout, cleavage only when you are putting those costumes on a lady with large breasts. Else it might look weird.

    Capelet refers to the pink jacket (or pink shrug) that women wear with their evening gown. It can be called independently from cape.

    Use "completely naked" to get rid of everything, and retain "white gloves", "white footwear", or "fontaine_mask", to create a defeated, naked version of fontaine. Add "torn pantyhose" to make it better.

    Version 1b addition:

    -Incorporated two additional variation to Fontaine's costume:

    1) Version with breast (clothing) cutout, revealing more cleavage, and see-through nipples.

    additional Prompt: lewdfonta1ne_dress, white dress, see-through

    2) Version with leotard, and more visible cameltoe/vulva.

    additional Prompt: lewdfonta1ne_dress, white leotard, highleg leotard

    Note that v1b, and v1 were trained independently, with a partial overlap in training set for the common costume components.

    v1b can technically reproduce the costume in v1, as long as you dont use the "lewdfonta1ne_dress" prompt. I trained both versions with the same activation token, so you can easily mix and match the components better by activating both LoRAs at the same time.

    ~~~~~~~~~~~~~~~~~~~~~~~~~~
    Recommended LoRA strength:

    Aurora v1. use 0.8 (0.5-0.9 are plausible ranges)

    v1b. use 0.8 (0.6-0.9 are plausible range dependent on checkpoints; higher values might lead to overbleeding of concepts, so beware)

    v1. use 0.5 (0.4-0.7 are plausible range dependent on checkpoints; higher values might lead to overbleeding of concepts, so beware)

    Main Prompt: gfontaine

    Proper:

    1girl, wearing gfonta1ne, [any size you like really] breasts, white dress, fonta1ne_mask, fonta1ne_hat, forehead jewel, jewelry, hat feather, cape, elbow gloves, thighhigh, holding, magical wand, white footwear, knee boots

    'sexy-upscale':

    1girl, wearing gfonta1ne, large breasts, white dress, fonta1ne_mask, fonta1ne_hat, forehead jewel, jewelry, hat feather, cape, elbow gloves, thighhigh, holding, magical wand, see-through, panties, open clothes, cleavage

    Removing fonta1ne_mask, takes away the domino mask.

    Removing fonta1ne_hat, takes away the white hat (though mightnot always work with high loRA strength).

    Removing holding, magical wand, removes the weapon she's holding, obviously..

    PS: Experiment with it and mix-and-match, each component is trained (in principle) to be detachable.

    Adetailer hints:

    You will need Adetailer to fix the faces, here's the prompt that I typically use:

    face of a mature woman, beautiful features, beautiful eyes, lipstick, fonta1ne_mask, fonta1ne_hat, forehead jewel, jewelry, hat feather, <lora here>

    Generally you might be able to get away to not using a separate prompt, but for full body shots or cowboy shots, you will need a reasonably high denoise ~0.45, to fix this accurately.

    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    Checkpoints that work well for this LoRA:

    1) Most checkpoints that I have tried (9 models) work very well with this. Including realistic ones.

    If you find one that cannot get it to work at all, let me know.

    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    Recommended approach: is to not use HiRes, but upscale it with img2img instead.

    Use a denoise strength of 0.25 -0.3 (anything higher will overcorrect face during the process), and turn off Adetailer. But up to you, HiRes works too.

    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    Issues:

    v1. The only issue I got here is with myself, that I forgot to add in the "subfolder" on lewd_fontaine into my training data, so now I will probably have to release an update/multi-concept version to spice things up. Stay tuned.

    v1b. The fonta1ne_hat prompt dont seem to work well, when used in combination with white leotard, i.e. a big white hat is generated instead of her usual hat. For the breast cutout variation, this is less of a problem. However, you can always re-seed to get a better generation, or artificially upweight "fonta1ne_hat". Perhaps I might re-train the model by adding more training images on that, to fix it at some point.

    Post-production edits/Reproducibility:

    Extensive use of regional/segment face A-detailer for the H-scenes.

    Happy prompting!!

    Description

    Epoch #12 chosen (Epoch #8 retained as 2nd best option)

    Trained on 261 images. No reported issues.

    Main prompt example: 1girl, f0nta1ne_4urora, fontaine_mask, fontaine_hat, capelet, white footwear, knee boots, cleavage cutout, cleavage, white gloves, white dress, belt, bowtie, <lora:Masked_Fontaine_aurora_v1_epoch_12:0.8>

    weapon-holding, magic:

    magical wand, holding wand, sword, holding weapon, special effect, magic

    See Tips in main model descrption page, for more.

    FAQ

    LORA
    SD 1.5

    Details

    Downloads
    379
    Platform
    CivitAI
    Platform Status
    Available
    Created
    5/13/2024
    Updated
    4/22/2026
    Deleted
    -
    Trigger Words:
    f0nta1ne_4urora
    1girl
    fontaine_mask
    fontaine_hat
    capelet
    white footwear
    knee boots
    cleavage cutout
    cleavage
    white gloves
    white dress
    belt
    bowtie

    Files

    Masked_Fontaine_aurora_v1_epoch_12.safetensors