CivArchive
    Babyfaced - PL_babyfaced_v2
    Preview 1
    Preview 2

    Use between 0.18-0.6 for subtle to more youthful facial features.

    Trained on 512x512, can use on 512x768 and 768x768.

    Use ADetailer for best results!

    --> Looks best between 0.3-0.5 for anime/stylized/cartoon checkpoints

    --> Looks best between 0.18-0.3 for real/semi-real checkpoints

    Only adjusts facial cuteness/softness. Change anything else with no issues.

    More details below>>>

    Full Training Details (things you should know):

    This is a model I trained mainly to improve the eyes and skin/makeup of a checkpoint I use. It's intended to give your characters a more youthful, full-faced, large/bright-eyed look. Looks good with real, semi-real, anime, cartoon/stylized, and 3D checkpoints. Experimented with probably 10+ different checkpoints and they all looked good.

    It's not a character so you can change eye color, body, hair color, and clothes with no issues while maintaining a cute/youthful face. Should work on different skin tones fine - most of the images from my dataset range in skin color from light to medium so there should be no bias.

    I recommend keeping strength between 0.18-0.6. I found that any less than that doesn't really improve eyes that much and any stronger than 0.6 starts to look overbaked. May still look good on anime checkpoints since it starts to look more exaggerated and painted.

    I trained on 512x512 but it looks great at 512x768 (portraits) and 768x768.

    I didn't train on full bodies as this is intended for faces, but I did include medium shots in my dataset so you should have no problem generating medium and full body shots. This lora is very flexible and will give you a thinner body if you ask it to, but it may bias towards large breasts if you don't tell it not to.

    Important Note:

    This was trained exclusively on bimboEngine_v10 (stylized semi-real/cartoon-ish checkpoint) and was not trained on or depicts any real people. I do NOT condone the use of this model for any abusive content. Sexual content is permitted so long as it does not involve minor characters.

    Personal note to users who made it this far:

    Please let me know if you have any issues! This is my first (released) lora but I love creating them in my off time - have been doing it for fun now for about 2 years and finally decided to start sharing them with you all.

    Hope you guys like it!

    Description

    Version 2.0 Notes

    • Designed to enhance facial features with a soft and cute look while maintaining full control over body, hair, outfits, and environment

    • Improves expressions, eye detail, and overall facial softness and harmony

    • Works well across a range of stylized, anime, and real checkpoints

    Recommended Strength:

    • 0.18-0.6 for most models

    • 0.18-0.3 for real checkpoints

    • 0.3-0.5 for stylized/cartoon/3D/anime

    Notes:

    • Higher strengths may exaggerate features or stylization

    • Best results with clean prompts and simple lighting

    • Can be combined with other Loras for styling and atmosphere

    • Should remain stable across all strengths (0.01-1) but looks best between 0.18-0.6

    • Trained on 512x512 and tested at CFG 7.5, sampling steps 30, karras schedule type, and using ADetailer

    LORA
    SD 1.5

    Details

    Downloads
    0
    Platform
    SeaArt
    Platform Status
    Available
    Created
    4/8/2026
    Updated
    4/8/2026
    Deleted
    -
    Trigger Words:
    babyfaced, 1girl, closeup, looking at viewer
    smile
    soft smile
    neutral expression
    surprised
    underwater
    side view
    three-quarter view

    Files

    Available On (1 platform)