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

NEXT PATREON EXCLUSIVE LORA (Patreon)

Published:
2023-08-11 17:32:18
Imported:
2026-06
Tags:

Poll

NEXT PATREON EXCLUSIVE LORA

  • New Character 210
  • New Style 323
  • New Object 86
  • Something else 55
  • 2023-08-11
  • 674 votes
{'title': 'NEXT PATREON EXCLUSIVE LORA', 'choices': [{'text': 'New Character', 'votes': 210}, {'text': 'New Style', 'votes': 323}, {'text': 'New Object', 'votes': 86}, {'text': 'Something else', 'votes': 55}], 'closes_at': None, 'created_at': datetime.datetime(2023, 8, 11, 17, 31, 4, tzinfo=datetime.timezone.utc), 'description': None, 'allow_multiple': False, 'total_votes': 674}

Content

Let me know what LORA you wanna see!

Comments

Michael

i chose new character but i am curious about animated characters / vector characters

Rai T00n

I know it's not popular but I wish to see object Lora someday :)

cblomert

would love to see an object lora as well - there is already too many ppl & style loras imo

ChopChop

i too want object lora, but soooo many ppl lora i want/need! lol

ProudChinaLover

Woudl you please teach how to use Dreambooth in Kohya?

ProudChinaLover

Plzzzzzzzzzzzzzzzzzzzzzzzzzz, I really would like to see how you teach us to use Dreambooth with SDXL.

Koudkunstje

I like to see a lora of a difficult object like for example a trench coat. SD(XL) has a difficult time with trench coats, random button placement, missing buttons, inconsistent buttons (form/size), buckles in general, multiple buckles on a belt, extra flaps, etc. How to train on a specific clothing piece?

Crux Riajuu

I've been looking for a model of kim jung gi art style but the only one there was for automatic 1111 was taken down - there is one for dreambooth but I only use 1111.

Roy Ding

there is not much on how to train a style online, please make a video on this.

Simon Kwan

Would be great to have an age slider to control and adjust the age appearance of any Human character but in lieu of that maybe start with age slider for a specific celebrity? Just to see how well it works?

James Clow

I'd love to know if a lora can be used to improve skin details. So many of the human models look like waxwork models.

Aitrepreneur

Yes I will make a dedicated video on this ;)

JoeyDotDigitize

I think model vs lora maybe? The main difference?

baran mengül

I thing sdxl checkpoint (style)-model / too difficult

baran mengül

Please do not have similar styles as in other channels, let it be a somewhat difficult art form, for example, "pixel art" actually increases the ironic resolution a bit, but our style requires some mastery

Al H

Something I'd like is styles in a certain era - for example, photos from the eighties look very different from contemporary digital images. Given your excellent training vid, I'm going to try that myself anyway, but would be interested in how you'd approach a style, rather than a person's likeness?

State of Mind

I would love to see a video tutorial on a LORA object, a specific t-shirt or some sunglasses. I haven't found anything top-notch on the internet about this topic, especially not up to the standards of the content handled here. Thanks in advance!

DanO..

How to do a unique costume would be great.

Aitrepreneur

Yes I'm interested in that too, I'll put that on my to do list

Aroen van Remmen

RunPod guide works perfect when following along the video, props to ya and thanks for your hard work. Just a tiny tiny piece of advice, when training the model, in stead of refreshing the log file every so often, just tail it. In your terminal just execute: `tail /path/to/file.log -f` And your terminal will 'tail follow' the file as it updates. Might be useful to you in the future.

Aitrepreneur

yes I saw someone else mention this, great trick!

Sugar Coat VFX Design

RunPod LORA training video is a masterpiece~! Would be great if you include a section about how to use a network drive to store files in Runpod so that i don't need to download things everytime.

Peruvian Dota

I already said it, but again thank you for the content.

Merrick Cox

I'm very curious if Style training can be done with regularization images that are identical to the training images except that an added layer is missing. I have paintings with and without a certain effect applied (don't guess!) and while I want to test this myself, I have respectable things I gotta' train first. :P

Aitrepreneur

you can just leave everything in the pod and stop it but you'll have to pay to keep those files in there (it's I think 0.02$ an hour)

Aitrepreneur

I need to do more testing on that, especially when it comes to regularization images too when training LORAs

s0up

I've been using adjustment loras for 1.5 like add detail, contrast, i've seen the clothing adjustment one, so I wanna see more of these. Idk what exactly tho

Naor Balori

i would very much love a spiderverse style!

Theron Holladay

Cyberpunk Styles for me. Hosting a new Cyberpunk Red campaign.

Andros Forever

A video detailing creating different types of styles

Steve

preferably something eclectic I'm interested in the difference between training a location example the Alamo (where its just a place that is timeless), Vs art style like Art Deco (has specific arrangement of things of shapes and colors), Vs a full genre like Alien westerns (both of the above with characters clothing and items) (I just picked things with A for examples plz don't use my examples they're arguably the worst). I'm also interested in crossing over lora's using the clips and model strength to coax out specific elements and when it actually makes sense to merge.

Therese

Hi, I'm a total newbie for training models so I hope I am in the right place to ask these questions. I watched your impressive video," ULTIMATE SDXL LORA Training! Get THE BEST RESULTS! " Thank you soooo much. It gave me the inspiration to try my first training.

Therese

I have run into a problem. At the end of the video, after "starting the training" you moved on to runpod and did not show where the files we were working on go, so I am totally lost as to where they are and what I should do with them. I have some questions perhaps you could help me with. Questions"Where are the models? What do I do with the models? Where do I find the file showing the progress of the training? What are the reg images used for? What are the npz files? Can you make a short video showing the last steps(above) for us newbies." Thanks you for the really valuable info in your video.

Aitrepreneur

It seems like you might want to either rewatch the video again I really explain and show everything, and also watch my stable diffusion sdxl installation guide (the video just before the lora training video) so that you know what to do with those models.

Therese

Hi, Thanks for taking your valuable time to send me the info

Michael Young

Would be amazing to get a better idea of how to train a Lora from our own artworks, to apply our own artistic style to ideas.

Aitrepreneur

yes it's gonna be the next video, coming soon ;)

Fabdream.ai (edited)

Comment edits

2026-09-08 20:00 would love to see, how to train a custom object, there is no tutorial anywhere about this
2023-08-24 22:38

Alex

Thanks for the tutorial! I vote for video about text2img of several persons at a picture, i.e. normal ones not mutants :) including "sixty-shmixty" content (nsfw scenes). ;) It seems SD is fine to generate a single person pictures, but really strugles to generate 2+ persons in action :D

Jalol Khamroev

I would love to see how you train some face (unpopular), I tried to train Lora with my face(there is no famous people who looks like me, so I took Chris Pine who had 14% similarity), so the results were kinda bad. Any suggestions?

Aitrepreneur

the perfect resemblance doesn't matter, it just needs to be enough as a base to make the training more efficient, what's more important is having a good selection of training images, they need to be of high quality, high resolution and display your face properly under different lighting an conditions

Aitrepreneur

yes the only real way of doing this is inpainting, this is the most efficient and practical way of doing this

Jalol Khamroev

Thank you so much! Appreciate your feedback and putting big efforts to make videos for us <3

Eric Bintner

Something else: quick overviews of colab and other tools. I get bored in 50min videos when I'm just looking for one piece of information.

The Stranger

Hello Ai Overlord! And fellow humans..can anyone tell me where I should start in all of this? I obviously was going to start in the beginning but I quickly saw some of these programs have been changed. Just looking for a sign that will allow me to build from the ground up and catch up. Thanks in advance.

Aitrepreneur

you can start by installing stable diffusion (watch my install video first) then go from there :)

baran mengül

You are the best.🙌 You work very hard, my gratitude. I couldn't wait for you, I couldn't wait for you. And thanks to you, I did 5 style studies. Now I get an error with 2200 normalized images and 110 working images. Cuda out of memory error. With 24GB VRAM (3090) in RunPod, which RunPod preset should I use for these tutorials (2200Picture/110 Picture)🤪? How do I resolve this error? thanks.

Aitrepreneur

sorry the style video took longer than expected, should arrive very soon probably this week, meanwhile if you get a cuda out of memory error on runpod here's why: long story short the runpod template was recently updated (like 2-3 days ago) to feature the auto1111 webui version 1.6 which apparently uses more vram being idle and because of that this uses too much vram on the gpu for a lora training, I contacted the creator of the template and he kindly created a new Application Manager to stop apps that you don't use, I will show how to use that in my next video but for now, if you want to train a lora on runpod you need to click on connect then click on port 8000 which will open the application manager, from there you will have the option to start or stop the A1111 (sd webui), kohya, and comfyui, if you want to train a lora with kohya, just click on stop for both a1111 and comfyui, then you can start the training. If you again want to use the SD webui after training, you can come back in that app manager and click on the start button for a1111, then wait a few seconds for it to start ;)

Duke Foxx

I would love to learn how to train clothes on SDXL

Aitrepreneur

yes it will be my next video after the style one ;)

EvanGuo

how to use prodigy