Fern from Frieren : Beyond Journey's End
UPDATE SDXL PONY VERSION, READ MORE BELOW!
SDXL Pony and SD 1.5 have the same trigger words.
Quite a little dataset of her using different outfit, please read more info below.
Strength : 1 - 0.6
Trigger Word :
fern \(sousou no frieren\), large breasts, long hair, purple hair, half updo, blunt bangs, purple eyesDefault Outfit :
open robe, long dressYou also might want to add :
boots, white pupils, huge breasts, hair ornamentKnown Issues
SDXL Pony
If you're having difficulty when prompting her to use a dress that have her chest a bit "bare", you can lower the LoRA strength or add "collarbone"/"cleavage" in the prompt
SD 1.5
Her hairstyle half updo sometimes showing the hair ornament wrong, some of the dataset in different angles are not having that, so, it's not perfect
Not tested in realistic model
Support & Commissions
If you enjoy my work, please consider supporting me on Ko-fi or commissioning a custom LoRA. Every bit helps me create more!
For commissions and more, visit My Website
To anyone considering reposting or reprinting this LoRA to other sites : Please respect my efforts and do not remove any links, including my ko-fi link, from this page. Thanks.
Description
Added version 0.5
FAQ
Comments (5)
Something that's been puzzling me it's that a lot of Lora's that I've seen recently on CivitAI require you to prompt the hair and eye color of the character (Which are basic characteristics the Lora should have learnt).
Could it be that the datasets have the eye color and hair colors tagged? That would be counterproductive. The intrinsic features of a character shouldn't be labeled on the training images, because that's what you want associated with the trigger word.
If you tag the hair color, eye color, haircut and any skin condition and accesory the character wears, you leave the Lora with only the face shape to learn.
Maybe you can try to prompt without those basic stuff?
Idk, I set the color and even Fern's blunt bangs so I can make her in blonde hair, swept bangs, different eye colors... just like someone would draw her as they like...
Oh, and the pruning is more like with her default outfit, on my side... so you don't need to prompt them that long like "long sleeves, dress, white dress, long dress, robe, black robe,"
Just "open robe, long dress", the rest are learned by the Lora
@Barjoe01 I did try at first without prompting the hair and eye color, and sure enough Fern came out with Blonde or orange hair. And sometimes because I prompted the black robe she had brunette hair.
I see your point about versatility, it's just adding modifiers to a character hasn't been a problem for me so far.
Regardless, with the full keyword list the Lora works fine, there's no issue with the accuracy of the character. But if we can get the same performance with a single keyword... I see it as an absolute win.
thats very debatable. depending on what the hair color or eye color is, you might be better off tagging it. if the color is closer to a common color actually having the color in dataset and using it as a prompt gives better eyes. if its a unique color you are better off not tag it.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.