CivArchive
    Aakiro 🍊 from Huggy - V01E07
    NSFW
    Preview undefined
    Preview undefined

    🍊 Who is Aakiro?

    Aakiro knew not the hour of its birth, nor the span of its existence. The essence of its being remained shrouded in the mists of time. It is said that life is but a fleeting moment, to be lived to its fullest. In the cherished companionship of its dearest human friend, Yumi Hiro (弓寛), Aakiro experienced the zenith of its days. The memory of their fateful encounter in the blossoming spring, on the twentieth day of March in the year 2025, would forever be etched in the tapestry of Aakiro's soul.

    📒 Version 2 is coming.

    • The upcoming version will be trained with twice the amount of training data as V1, with improved details on claws, paw pads, and eyes. Additionally, it will work more smoothly with extra characters.

    • The "Slow Baking" training technique will enhance generalization, making it easier to work with other LoRAs and base models.

    • This new version will be trained using the Illustrious base model.

    📒 Recommended settings:

    Use CFG Scale = 6

    Steps > 33

    Use this lora to improve overall quality without using triggering word

    📒 Useful Tips:

    Use V01E04 for PVC figure rendering

    Use V01E06/08 for anime rendering

    Use V01E09 for plush rendering (as you see in the music video)

    Under V01E06 section, you will see 35 sample images from same prompt generated with different checkpoints for comparing purpose, and here is the result.

    I'm currently testing different epoch versions, having randomly selected V01E08 to start, alongside my previous Loras like "Buddy" and "Morishima Hodaka." So far, I'm pleased with the results. It does rely on a negative prompt to fix the paw pad, so feel free to check out my showcase samples for reference.

    Please note, this is still a work in progress as I'm trying to find the best epoch version out of the six files. In the meantime, I've already started preparing for the second version.

    Since the start of this year, I've trained nearly 40 Loras. To help speed up the process, I’ve been using AI tools to create custom programmes for resource preparation. I also manually edit every single image using brush and eraser tools to ensure perfect detail. Writing training tags for every image is another step I don't overlook, as I’ve found it’s crucial for quality.

    Training a good model still feels a bit like metaphysics to me—I'm constantly figuring out the right balance of training resources, repetition numbers, and training steps. It's tough to nail down the perfect values because they’re dynamic and depend on the complexity of the style.

    Though Aakiro was trained with tag 1boy, it's considered as a gender-neutral spiritual animal without genitals.

    All my loras are Civitai exclusive, please let me know if you see it on other platforms.

    Description

    Resolution 1408 x 1408

    Max Steps 840

    Image Count 32

    Repeat 13

    FAQ

    Comments (17)

    AshWasTakenDec 25, 2024· 14 reactions
    CivitAI

    uh yes a lora from the future.

    gynoidnekoJan 1, 2025· 3 reactions

    Seriously. This gets unfairly pushed to the top of the most recent models list and will be like that for months before we have to stop looking at it. Not fair to the other people posting models.

    LusciousFawkesDec 27, 2024· 4 reactions
    CivitAI

    This spreadsheet is great, every LORA needs this. Nice work.

    huggy
    Author
    Dec 27, 2024

    cheers!

    StinkekDec 29, 2024
    CivitAI

    "I just find out this lora was locked by moderator for any NSFW content (for what?), which explained the weird setting changes behind my back..."

    It's probably to prevent generation of sh*ta. Any cute/petite character may be considered a child.

    huggy
    Author
    Dec 29, 2024

    since when puppy / pussy needs censorships, what next then, put clothes on all animals in real world? BTW this lora was not trained from minor furry. I simply used prompt like young / size related tag to make it cute, and this technique can be applied to any lora / checkpoint.

    StinkekDec 29, 2024

    @huggy Civitai is very strict when it comes to young-looking characters. Tags like "young, cute, petite" are associated with minors. Anything that looks like an even slightly lewd depiction of a minor may be flagged as "child abuse and exploitation". This applies to anthropomorphic animals too, even if they're just sitting with legs closed and wearing stockings with no pants on. Guess why I'm being oddly specific...
    Because the point of this LoRA is to make a cute character, who may be easily perceived as a child (purely based on looks, lore is irrelevant), they likely disabled NSFW generation because generating shota would be too easy otherwise.
    Basically anything "young, cute, petite" and NSFW should not be mixed because it may be perceived as "child abuse and exploitation" regardless of intent.

    huggy
    Author
    Dec 29, 2024· 1 reaction

    @Stinkek It's a fallback of censorship, a lazy approach to management, and misuse of regulation, especially on a creative platform. You can find countless videos like 'kids kiss animals' on TK, yet similar artworks are being deleted. I'm seriously annoyed with this move.

    StinkekDec 29, 2024

    @huggy I don't know what TK stands for. I assume what you described means a situation where a child expresses affection (obviously innocent) for a regular animal, I don't see what it has to do with a furry LoRA.

    I just decided to use Pixiv for anything I think wouldn't pass on Civitai, or it unexpectedly got taken down (another case: it doesn't say anywhere in the ToS that I can't post characters being hung with a noose, I just found out about it after speaking to a mod).

    huggy
    Author
    Jan 6, 2025

    @Stinkek Take a closer look at our “Guardians” like a user called "DaOneWhoDontSnooze", and all his pedo works, you'll understand why this site is ruined with these “Guardians”.

    StinkekJan 6, 2025

    @huggy Furries are admittedly harder to label as adult or loli/shota, a picture of mine got slammed for much less than what this user makes, I've been paranoid ever since and I generate furries with big muscles or other features just to steer away from the grey area, even if I don't quite enjoy what I do. I want to retrain a furry LoRA of mine sooner than previously anticipated for this reason.

    StinkekJan 6, 2025

    @huggy Also I wouldn't be putting that bold text for the public to see if I were you, After I questioned the removal of that one picture in an "article" (which I deleted shortly afterwards), I found the users here are too... cold-blooded for me to be comfortable around. I decided that whatever complaints I have about this site, I WON'T be making them on here so I don't get humiliated again. I wish whatever you're doing doesn't result in drama.

    huggy
    Author
    Jan 6, 2025

    @Stinkek I'm sorry to hear that, I know we are all victims, my point is they are playing with the rules to get survived here by pushing others aside like squid game, and the site owners properly enjoy watching this

    accrossDec 30, 2024· 7 reactions
    CivitAI

    lovely character, please make a flux version LoRA Nijijourney V5 Furry Style

    huggy
    Author
    Dec 30, 2024

    My last flux lora was a disaster lol, if you wanna train it, I can send you the training file

    montacDec 30, 2024· 6 reactions
    CivitAI

    just making a cosplay costume out of it for next comic con, should it has 3 or 4 toes :)

    ArtictJan 3, 2025· 5 reactions
    CivitAI

    Omg, this is TOO cute, it should be illegal 😩😭 I can totally see this turning into a whole manga series, like, it’s got SO much potential 😍✨

    LORA
    Pony

    Details

    Downloads
    26
    Platform
    CivitAI
    Platform Status
    Available
    Created
    12/19/2024
    Updated
    5/19/2026
    Deleted
    -
    Trigger Words:
    Aakiro,source_furry,grey fur,orange ear,

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.