Mangamaster Style - Pony / Illustrious

    Mangamaster style V4_epoch_8.safetensors LORA / Illustrious

    Model Information
    Model Name
    Mangamaster Style - Pony / Illustrious
    Version
    Illustrious
    Creator
    Ava_Choco
    Size
    217.88 MB
    Downloads
    4,203
    Trigger
    Mangamast3r,
    Torrent Details
    BTIH
    5460E9C19DE3EC07F837805D1240CA1F19FCA353
    BTMH
    566AEB8BEBE9FBE876BF5BCFEE1BC6F78C84EA3E46EEDA975D19EFA1A22ADA75
    SHA256
    2FE429D3D5CC7C9691917FA99ED5D85C26FF382C61FE9EAD0D12F3951A017EF7
    Upload Date
    about a year ago
    Uploader
    CivitasBay.org
    Status
    2 Seeders
    0 Peers
    Info

    The style of Mangamaster.

    I made this lora to try and reproduce the style of Mangamaster.

    I added a new version trained of Illustrious 2+. I also changed the activation tag to mangamaster

    This LoRa is trained on Pony V6 and Illustrious 0.1, but I find the results in Autismmix (Pony) and Obsession (Illustrious) more to my liking. This lora works great in both checkpoints!

    Both image sets were made with Addetailer extension enabled and using a face and eye model (and some other were needed).

    Trigger: Mangamast3r,

    Advised strength: 0.7-0.8

    Prompting Illustrious

    For all versions I recommend you use: masterpiece, best quality, amazing quality, very aesthetic, high resolution, in your prompt and I myself am using the following negative prompt: graphic, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, very displeasing, lowres, (worst quality, bad quality:1.2), bad anatomy, jpeg artifacts, signature, watermark, username, signature, simple background, conjoined,ai-generated, muscular, 3D, polaroid,

    Prompting Pony

    For all version I recommend you use: score_9, score_8_up, score_7_up, score_6_up, in your prompt and score_4, score_5, censored, mutated, bad hands, deformed hands, missing fingers, multiple fingers, multiple digits, extra digits, in your negatives. It will greatly improve the results.


    Version 1:
    This is one of the first lora I have made and trained on 6000 steps and 10 epochs. The dataset was a of normal quality. The images created by this style have a smooth feel to it. I used the addetailer extension to help with the faces and eyes

    Version 2:
    Trained on a dataset double that of version 1, but with a lower step count (4100 steps to be precise). It's feel's stronger in style then version 1. I used the addetailer extension to help with the faces and eyes

    Version 3:
    Trained on a dataset of 30 images of high resolution. The number of steps has been increased to 4500. The style is very strong so giving it a lower weight might help. This version produces sharp lines. Also the trigger word got changed to Mangamast3r

    Thank you for reading and have fun!

    Gallery
    This model contains NSFW content. Click to show gallery.
    Trackers
    • udp://tracker.opentrackr.org:1337/announce
    • udp://open.demonii.com:1337/announce
    • udp://open.stealth.si:80/announce
    • udp://tracker.torrent.eu.org:451/announce
    • udp://tracker.qu.ax:6969/announce
    • udp://tracker.bittor.pw:1337/announce