Each version is a unique mix and will give you a different consistent face, they are not building upon the last.
I updated the names of the files. The names might not match older gens but they are the same.
Please share your gens below and rate/react, it means a lot and helps me know which y’all like and dislike. I love seeing what is made with them, even if they are at a low weight.
I am bad at coming up with names so I am just doing alphabetical for now. If y’all think of a name that fits any of them please let me know!
A group of loras I have slowly built up as I enjoy having consistent faces across gens Each uses an assortment of faces from different sources. I will be uploading as I make more! Hope you enjoy
Description
This is a zip file of all of the nobodies I have posted
Contains: Azalea, Bryony, Camellia, Dalia, Ellen, Fiona, Gardenia, Helena, Iris, Jasmine, Kelly, Lily, Magnolia, Nadine, Olivia, Primrose, Quinn, Rose, Sasha, Tulip, Uva, Violet, Willow, Xena, Yuzu, Zinnia.
FAQ
Comments (26)
Hey just wanted to say that these have worked really well for me when I need a female character with continuity being important. They have all worked really well with all kinds of checkpoints and other injectors/networks. Great work on them, actually believable human models is hard to come by in the SD world haha.
I am glad they are working well! Would love to see some of the gens you have made with them.
I love all of them, Was wondering if we could make requests or commissions? looking for a goth one haha
I will sure consider that! But it might take a bit to do. Would love to see some of your gens you have done with them so far.
@ShubbleLover is there a way to contact you? either discord or email?
@confettibearr sure. My disc is shubblelover
@ShubbleLover Sent a friend request, whenever you get a chance to respond would be great <3
I really like it that you made an all in one pack, that makes it so much easier to D/L, wish more ppl did that.
they don't look like nobodies to me! 😍😍😍😍😍😍😍😍😍😍😍😍😍😍😍😍
.
Can we expect to see this Lora trained for SDXL1.0?
I can look into it.
Thanks. I like Dalia quite much :)
Here you go, I have started training them. https://civitai.com/models/160073/sl-assorted-female-nobodies-xl
Hi !
First of all thank you for sharing and really well done !
I needed consistent faces for a comic project and this helped me a lot !
Could you explain to me briefly how you create these loras ?
Any plan of creating a paid course ?
You could create like a mini ebook where you explain how you've done it , and put it for sale like for 20/30 dollars im sure people would buy it !
(i would but it for sure )
Thank you for the kind words and I am glad you enjoy them! I look forward to seeing what you make with them.
As for how I make them, I use Kohya's google collab. I run Auto1111 on AMD so training locally is a pain. As for a paid course, I do not have interest in that. I will happily share and walk someone through how to use Kohya.
@ShubbleLover Yeah i know kohya and used it too. But what i wonder is how you create a face ? what's the dataset on the lora you train ? how do you get this dataset ?
@ArtificialBeauties I use faces that are similar in shape and have similar facial features. for the datasets, it is just gathering assorted photos. It works best when the face shape and features are similar through the different faces in the mix. I try to have about 5 faces in a mix. 25% full body, 25% HD up close headshots, and 50% half body. 10% behind, 20% from the side, 70% from the front. I try to not use less than 25 photos and I have used as much as 500.
@ShubbleLover ah i see , thanks for the information, but want i dont understand are the faces, lets say for the lora " Azalea" , so you searched women with faces , you took like 5 different women , with different faces , and created a lora with them ?? so let's say i find 5 different women ( or men ) , i take 6 pictures of each , so i have 30 pics , and i create a lora with these ??
@ArtificialBeauties pretty much! For Azalea I focused on French women. Sometimes I do redheads, sometimes I focus on jawlines.
@ShubbleLover ok i see, and for the tagging , you just give a random name in the txt files to your subject ??
thanks for this i gonna try out to create a male for training
@ArtificialBeauties yep! The names are random.
@ShubbleLover thanks a lot , gonna try this out !!
thanks a lot for the zip of all the models, maybe the other creators don't do it because they want to get like on each model, but I think not everyone wants every one of them, and you can ask the people downloading the zip to drop a like on other models to,
I will like every one of them because I like to support
Why is this 26 LoRAs? 😅 A single character can be done better in a text embedding, and a single LoRA can do 26+ characters.
It’s because I was learning how to do them. I have an amd system so I was limited on what I could do at the time.