CivArchive
    Harry Potter Characters Pack LoRA | Hermione Granger, Draco Malfoy - PRP v1
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined

    Harry Potter Characters Pack LoRA 💥

    This LoRA is trained on a scene image dataset from the Harry Potter series, from The Philosopher's Stone to The Deathly Hallows. The idea is I want to have a model that can remember all the main characters and able to recognize the different between 1st year characters and 7th year characters. However, the result is not so successful. I still want to share some part of this project.

    Usage Tips:

    • This version only with my PrincessRealitic Ponly model. The version that work with Pony Diffusion V6 XL or other Pony-based models are not good (still in development).

    • Trigger Words for Characters: harry potter, hermione granger, ron weasley, draco malfoy,

      • Some more Trigger Words (that used but not show well result)ginny weasley, neville longbottom, dudley dursley,cho chang, cedric diggory, fleur delacour,viktor krum, luna lovegood,george fed weasley,

      • The model is also trained with year Trigger Words but not seem to follow the tags.: year_1,year_2,year_3,year_4,year_5,year_6,year_7,

    • Sampling Tips: For best results, I recommend using this model with ADetailer

    Description

    FAQ

    Comments (5)

    noyartMar 22, 2025
    CivitAI

    Any plans to release separate models of each character? :)

    titansteng
    Author
    Mar 22, 2025· 2 reactions

    Not yet, not soon. I still have to find some good parameter setting for the training. Before moving to any next project.

    ravemry9Mar 26, 2025
    CivitAI

    A pity you say this project didnt work out well, your models are usually excellent.

    Any idea what went wrong? especially with your idea of year tags (great idea, more people should try it!) -- maybe SDXL can't differentiate between years since the faces are too similar?

    titansteng
    Author
    Mar 28, 2025· 2 reactions

    My work are not usually excellent. There are a lot of bad works that I don't want to share. In the case of this model, there are people DM to me that they would like to have harry potter's lora which I have.

    I don't know what went wrong. Maybe the size of dataset is too small for the task. Recently, I have tried to reduce the training cost so I can test more ideas. This project as well also have a limit in dataset size. I have remove some hogwarts professors and death eaters (also fantastic beasts characters). I will come back to this project soon. Maybe I'll try only harry from different years. I also want to know what wrong.

    ravemry9Mar 28, 2025

    @titansteng OK, thanks for clarifying.

    Keep up the good work, so much of your output is very creative and very useful, I really enjoy it.

    LORA
    Pony

    Details

    Downloads
    437
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    3/4/2025
    Updated
    5/14/2026
    Deleted
    5/23/2025
    Trigger Words:
    harry potter
    hermione granger
    ron weasley
    draco malfoy

    Files

    Harry_Potter_PRP_V2.safetensors

    Mirrors

    TensorFiles (1 mirrors)

    imgIII.zip

    Mirrors

    HuggingFace (1 mirrors)
    CivitAI (1 mirrors)