CivArchive
    F.1_Generate Abstract Headshots - v2.12
    Preview 74111722Preview 74111861Preview 74111863Preview 74111864Preview 74111862Preview 74111936Preview 74111933Preview 74111932Preview 74113831

    A model that generates ugly avatars for work, which is perfect for expressing your good mental state. After all, who wouldn't go crazy at work? Invite your friends to use it too!

    Model Description: The training dataset contains a large amount of high-resolution materials. It has good versatility, and we welcome everyone to test it.

    Trigger word: Ugly avatar.

    Recommended graphic size: 1024*1024

    Sampling algorithm (Sampler): Euler.

    Scheduler: simple, normal, or bata.

    Sample frequency: About 25 steps.

    Recommended model weight: 0.8-1.2

    CFG: 3.5

    High-definition restoration: Using 1k as an avatar is enough. If you want to enlarge it, we recommend the following enlargement algorithms: R-ESRGAN_4x+Anime6B, 4x-AnimeSharp.

    Redrawing amplitude: about 0.3 (the smaller it is, the more it looks like the original image, and the larger it is, the more you can use your imagination).

    Reference keyword examples (click the images on the homepage and the return images to see the accompanying keywords):

    Due to being too abstract, testing hundreds of draws is still an abstraction, even more abstract than the original dataset. It's better to use trigger words for random draws...

    If you really want to use words, you can describe colors, hairstyles, face shapes, etc. Just have fun with it.

    Here are some test data:

    [Disclaimer]

    1. In principle, this model is only for communication and learning, and the model and generated images cannot be used for commercial purposes. If you need to use it commercially, you need to make enough manual modifications to the work to make it be judged as your personal work.

    2. Users shall use the content generated based on this model in compliance with their local laws and regulations and these Terms of Service. Any intellectual property disputes arising from the use of this content shall be dealt with by the users themselves.

    3. Without the author's permission, you are not allowed to host this model and its derivative versions on other service platforms or integrate them into applications and websites.

    4. When generating and posting pictures on social media, you must indicate the reference model and the author.

    Description

    LORA
    Flux.1 D

    Details

    Downloads
    60
    Platform
    CivitAI
    Platform Status
    Available
    Created
    5/4/2025
    Updated
    9/27/2025
    Deleted
    -
    Trigger Words:
    Ugly avatar

    Files

    F.1_抽象头像生成_v2.12.safetensors