Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

4 LORA MODELS (Patreon)

Published:
2023-08-11 17:24:12
Edited:
2023-08-11 20:21:45
Imported:
2026-06
Tags:

Content

I've crafted four unique LORA models based on some of the most iconic figures in the entertainment industry.

The Collection features:

Margot Robbie

Cillian Murphy

Billie Eilish

Milly Alcock

What's Inside Each Folder?

With each LORA model, you're getting a complete behind-the-scenes look into the training process:

The Model

Training Images

Image Captions

LINK TO DOWNLOAD (PATREON doesn't support files over 200mb....) : https://bit.ly/4lora-modelsor mega link (copy paste that full URL, patreon has trouble displaying the link) https://mega.nz/file/uU0BhZRD#RdahSuvEKLpAvjomftmZQL5pu_dQag0b-iBySTGaYCU

password: aitrepreneur

Let me know if you have a problem with the download!

As always, supporting me on Patreon allows me to keep creating helpful resources like this for the AI art community. Thank you for your support - now go train some awesome LORAs!

Files

Comments

ChopChop

Hi, the decryption key is not working friend.

Aitrepreneur

are you sure? It's working on my side? I think patreon is just badly displaying the URL, you need to copy and paste it, I will try to find a better solution soon

ChopChop

Ive tried a few times, with and without spaces at the end and beginning of the key. is it this? -- aitrepreneur

ChopChop

Thank you! I see now!

Aitrepreneur

I also added another link for the download, it should work fine now ;)

amazingghost

Getting this error trying to run the billie eilish model "A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check." using --disable-nan-check gives me black pictures

Aitrepreneur

I have the same error each time I run auto1111, it's a bug. I usually just generate a random picture 512x512 resolution without any lora, then I can generate the images I want

Fabdream.ai (edited)

Comment edits

2026-09-08 20:00 The zip file seam to be broken, impossible to extract the data
2023-08-24 22:42

Aitrepreneur

hey yeah I just tried it works for me, try maybe using 7zip or winrar to extract not sure what your are using (I'm using winrar)

Ken Biles

Yeah, you can't right-click and extract all. You have to use 7zips or WinZip.

POLYHEDRON_AI

Ah! You get this big files sizes also ... hm. Again: Do they have to be that big? :D Cheers!

Aitrepreneur

I prefer quality over file size but as I said you can just use a smaller network rank for smaller file size in exchange for a small quality degradation

Zen Crypto

I'm sorry. I'm truly confused. Are these LoRas or Checkpoint? My Stable Diffusion doesn't detect this as LoRa but if I add it to the Checkpoints folder, it is detected. But the model isn't accurate. It can't create Billie Elish. A clarification would be appreciated.

Aitrepreneur

It is a LORA, you add it to the Lora folder, it should appear automatically, if not refresh the list and make sure you select an SDXL checkpoint

Aitrepreneur

check your nvidia drivers, I also know a lot of people having issues with 4090 taking a lot time to train, that's a problem with kohya unfortunately

jamal

once i extract the zip folder it just gives me abunch of empty folder with names but nothing in them

Damien Benoit

Hey i tried the mega.nz link and it asks for the password but says that "aitrepreneur" is not a valid password. Did you change it or did I just do something wrong?

Aitrepreneur

no it should work, don't forget to use winrar for the archive extraction though

Damien Benoit

That Mega.nz link file, does not let me download it. I click it and it asks for a decryption key. I thought that was the password "aitrepreneur" obviously not, and more than likely its your zip file password but I still am not able to download whatever that file is without whatever the decryption key is. I didn't see that on this page anywhere though!

Damien Benoit

OMG, man i swear I lose access to half the things that I read when i read it. Im sorry, I just dont pick up on everything I read like all the time. I had to read it a few times to understand it wanted me to copy and paste that link to get the file

Pauline Fannes

I downloaded the files and placed them in model > lora and they appear fine on SDXL but when I try to use them, it doesn't work... What should I do to make it work?

Aitrepreneur

you need to select them (the lora name appears then in the prompts) + use the name trigger in the prompt to use them, check my lora training video, I show how to use them at the end

srob

These are great --do you keep the same JSON config file for each one? I have a hard time referencing the older A1111 videos you make and trying to pause to see the differences in the UI from today. Am I maybe just missing where you are sharing them? I'm new but really enjoying your content! Nice work!

Aitrepreneur

yes more or less these lora were all made with the same parameters

srob

Yes, I think that is what I'm trying to figure out. I see this set: https://www.patreon.com/posts/object-lora-91936208 --but those are more for objects, or runpod no? Also a bit older, and there are new settings (including presets it appears) in the new a1111, and your loras are excellent, but would love to just have the latest export of a JSON training any human, really, instead of building my own JSON that seems to be working, but feel like I could have better settings. Can you share your latest favorite settings for human generated Loras? Any even any tips on the best models to use to train on? Great work so far! Glad I joined!

Aitrepreneur

yes those presets are for objects, I made separate training videos for people, style, clothes and objects. Those json preset are optimized for pretty much anything you may want to train but the most important part is the dataset no matter what you train

Bradly

I just finished watching ULTIMATE SDXL LORA Training! Get THE BEST RESULTS! is there a way to train using a celebrity but then when you prompt you don't have to use "Jessica Alba woman"? But rather the persons name?

Aitrepreneur

no that's the whole point of training using the celebrity name

Tomas Rampas

I tried to download from Mega files but it says the key is wrong.

Antonio B.

why the lora weights are so heavy? Usually lora is under 100mb

Antonio B.

how many repeats for each image?