Aitrepreneur patreon

NEXT PATREON EXCLUSIVE LORA

🕑 Added 2023-10-12 16:22:32 +0000 UTC

Comments

Aitrepreneur

I already made a video on lora style training: https://youtu.be/1BCYdd9r1To the principle always stays the same

gabbatv

Hi! What would I use to create a LoRA for my own Artworks? The shaopes are never always the same like a LoRA object, but they do have a style like a LorA style, i'm not really sure what to choose! : )

Liam

Can you make a lora of Elle Fanning?

Víctor Ahumada

It worked perfect, thank you!

Aitrepreneur

yes but the best way to do this is with inpainting

Víctor Ahumada

Hello, I just saw your object training video in SD. I have a question. Can you combine an already created character LORA with an object LORA? Like for example adding a ring to an existing LORA?

Stylised Cartoon Character

Cartoon character

ZenzeosArt

Hairstyles

ComfyUI tutorial with instant LoRA please

Controlnet for SDXL tutorial for AUTO1111 please

Jan-Michael Losada

A handheld object like a sword would be sweet.

I think 'concepts' would be interesting. Things like lighting, extra detail, de-aging, fixing hands etc...

Phillip Gimmi

yes or a particular mask

Aitrepreneur

send me some more info in dm

Paula Kühn

I tried to train a specific bag in a shape of a vegetable and somehow this didn’t work at all, would love to see how you would do it! Character even if they are very abstract works great for me (sdxl rtx 4090)

Aitrepreneur

not really, reg are optional, they can make the model better but you can train one without them

I see. Then you have to get a bunch of Regularization images I guess?

Aitrepreneur

well it's the same principle as when you want to train a person, except you use the race of dog or cat for the token. So if your dog is a german shepherd, you will simply put "german shepherd" for the instance prompt and "dog" for the class prompt, simple as that :)

Would be nice to learn how to train a Lora on a specific dog or cat so you can put your animal in any picture / settings you want :)

Aitrepreneur

it doesn't work like that, correct hand generation is a big issue for most ai models, training a small lora is not gonna fix that, we'll probably have to wait for the next few SD base models for it to improve

Aitrepreneur

object training is probably gonna be my next video :)

Vector design (Logo style)

Hadriex

Or hand an object such as a handbag from one character to another and have that second character carry it in future generations.

George Gostyshev

Any chances to get a proper training LORA for mutated hands, legs and etc? It'll be sooo much appreciated by whole community of SDXL users.

For me it would be very interesting to find out how to train objects that interact with a model, e.g. a chair you are sitting on. Or how to add more datasets to an already trained Lora, i.e. how to resume the training. Thank you.

cblomert

oh but for this current poll: would love a pose lora of people taking a selfie :D the best would be a combination of a 3rd person view and the actual selfie that was taken .. but that is quite a lot to ask haha

cblomert

to be honest - i don't feel i need the exclusive loras - i don't know how the other patreons feel, but if it takes a bunch of time and effort - i would say skip them- and rather poll us for general topics for you to explore and explain to us :D for one: i would love to learn a bit more about the current state of 'local' LLMs and how to finetune them much love <3

Objects is so essential, we have so many styles already

CJ

poses

Aitrepreneur

A lora is not gonna fix that, it's an issue with most Ai models unfortunately :/

Objects like holding items would be verrryy useful. Currently making a space marine Lora but hands and weapons often get messed up.

shen oracle

pose

vector design!

Hands or teeth. It's amazing how hard it is to get either right.

Mahmoud Alhashim

architecture

in the spirit of Halloween which approach would be best to make a Lora to apply make up effects on someone like witches noses, vampire teeth, elf ears, etc.?