CivArchive
    Math Pellerin SDXL LORA - HIGH 1.0
    Preview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefined

    Mathieu Pellerin LORA Model

    By monamibob

    With Mathieu's permission I made this model using 25 or so images collected from the official instagram page of Mathieu Pellerin.

    I followed the recommendation from Aitrepreneur regarding SDXL lora trainings and used a known person as starting point for the training.

    In the case of Math Pellerin, the starbyface.com website gave a lot of different results but Jason Ritter came back as first or second result a few times so I used this person as the stable diffusion training starting point.

    Because of this The trigger for this lora is “mathp jason ritter”. Make sure to start all your prompts with this token!

    LORA training produces arrays of loras with more or less training and choosing the best model is a difficult and subjective task.

    I tried many prompts/styles/settings/content with multiple revisions of the models and in the end I selected 2 versions for publishing:

    1- mathp_jason_ritter_high
    a large model that retains high precision and was submitted to a lot of training. This huge file can produce the most realistic reproductions of the model but is less flexible and less able to create the subject in different styles or poses. If the intention is to make images that look like real life, photographs, portraits etc, this is the best model to use.

    2- mathp_jason_ritter_low
    A small model that doesn’t retain all the precision and didn’t go through as much training. This model is much smaller (220mb), less trained and a lot more flexible while still able to create strong likeness. I had an easier time using this LORA version with other LORAs. If you intend to create more abstract pictures or images with significantly different styles, this is the best version to use.

    For convenience I’ve renamed these models for publication. You may see them referenced as something else in the images published.

    mathp-jason_ritter-000001 --> mathp_jason_ritter_high

    last-000004.safetensors --> mathp_jason_ritter_low

    Model of Math Pellerin. Socials:

    https://instagram.com/mathieupellerin._/

    https://twitch.tv/mathpellerin

    https://youtube.com/@mathpellerin

    https://tiktok.com/@mathieupellerin._

    Model trained by monamibob. Socials:

    https://clipbattles.io

    https://youtube.com/@monamibob

    https://tiktok.com/@monamibob

    Thanks to AItrepreneur on youtube for the great video that helped me a lot understand the basics and get started with good parameters


    And of course many thanks to Mathieu Pellerin for giving me permission to use his likeness and images!

    Description

    High Training, High Precision version.

    Large file.

    LORA
    SDXL 1.0

    Details

    Downloads
    52
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    2/5/2024
    Updated
    7/7/2025
    Deleted
    5/23/2025
    Trigger Words:
    "mathp jason ritter"

    Files

    mathp_jason_ritter_high.safetensors

    Mirrors