illustrator Marcel Marlier.
thank @settima_ai 's suggestion.
Description
suggested weight 0.8-1
FAQ
Comments (9)
Hey, Your LORA is a success! : D
(...but where is gone my post with the image of the brides I made? )
yes, I saw your post, innocent one, two girls with wedding dress. why...?
@yuberkley in the "orangesClockWork" checkpoint place, the author posted a portrait with a child plus a grandparent to show the good age range of his checkpoint. CivitAI blocked the entire checkpoint for days and the author had to change the image, showing now two older people! Be very cautios in future because there is a sort of "witch hunt" around! Here you can see the image, how is NOW: https://civitai.com/images/614739?modelVersionId=55190&prioritizedUserIds=381569&period=AllTime&sort=Most+Reactions&limit=20
@yuberkleyI can just image the idea of a "too young bride" was not politically correct. Who knows...
@settima_ai wow, you are really the wikipedia of CIVITAI. I guess they check the prompts first, if fail they tag the pictures with AI, and "hunt the witch" with those tags.
Thank you so much for this, you STAR. Have been looking for a while for anything Marcel Marlier ... and you did it! Now off to create new adventures for Martine and Debbie
hello, tried this. changed to lyriel + prompts yet my creation has disfigured arms and legs, not as clear either.
NOT GETTING THE RIGHT RESULTS? TRY THIS
for absolute beginners like me:
1)download the same checkpoint file as the lora has used.
2)copy all the generational details as well
3)instead of generating a default size of the picture, 500x500 go for 768x768. (this will immensely improve all of your lora disfiguration issues.
4)click generate
5)now click Hires fix with latent(bicubic antialiased) using a denoising strength around 0.45.
6)generate again, amazing resutls just like in the examples show.
(this might be a lengthy explanation, there might be a short-cut to this, but i;ve shared anyways)
thank you for your summary, totally agree.
usually, when I test my trained Lora, I will use the prompts derived from the sample images used for training. So just copy the prompts and other details of the 'official" image is the safest way. after that , you can replace , add or remove prompts, or change base model to unleash your imagination.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.