1-Click FLUX LORA Training

12.71k views2505 WordsCopy TextShare
Olivio Sarikas
In this video, I’ll show you how to train the Flux Model LoRa for stunning, hyper-realistic image ge...
Video Transcript:
this oneclick flux Laura training is stunning hello my friends and how are you doing today I want to show you something that I'm pretty proud of because I worked very hard for it and I also want to show you some amazing tricks throughout this video on how you can get way more out of your flux Laura but first of all this is a sponsored video but I had to do some hard slaping to get the functions onto the page I wanted to have for you this includes that you can download the Laura that you have
trained the pric is a lot lower that you have control net for flux that you have in painting for flux all that good stuff that I said it has to be part of that page so I'm going to show you all of that and it works fantastic let's look at some examples first so here you have some really cool results this is not a training image this is what has been created by the flux low trained on my face and I've never seen anything that looks as much as me as these images here we have
another example and I could basically become an AI influencer pretending to travel the world and do the most adventurous stuff because it just looks like me and it just looks like a photo here's another example in the mountains check that out this even has this kind of wave that my beard is doing when it's getting longer and also the jacket that I often wear I'm really impressed on how much that looks like me here's another one the Arctic Explorer pretty cool image really like that but we also have here the business version I'm never going
to look like that to be honest I don't like to wear suits but it looks pretty good to be honest if this is the alternative reality of me I'm okay with that look here we have also the server dude I'm kind of missing out actually on a lot of cool action and by the way I'm super surprised it's actually got my chest here right even though it doesn't even know what that looks like again this really looks like a GoPro video of me doing some amazing surfing and here we have artistic style images also of
me this is with some tricks that I want to show you on how you get a very flexible flux Laura the problem with flux luras is they are very sticky to the images you have trained them on they're very precise in what they do but they are not very Broad in the range of what they can create so I figured out a trick for you how you can get a very wide range that is pretty amazing so here we have some other examples this one is a little bit more goofy but I'm super okay with
that because it's a pixel style it has to be cute and here we have a pencil sketch or a line art drawing that also looks really fantastic I love the artistic expression here and next we have here two different images two different methods that I also want to show you and how you can create results like that now the left version looks more like a 3D model with the actual 3D kind of style the right one has more of a photographic style even though it is 3D especially when you look at the beard hair detail
and here we also have another artistic style that looks pretty cool and actually like a painting again also from that Laura you can bu the way also train Styles so here I have a comparison between the input designs that I downloaded from free pick and the images generated by that flux Laura again pretty stunning results and super easy now in case you're wondering why I am so impressed with all of that well this is is a comparison between my real face and the image created by the flux Laura in the middle you can see here
this is the photo that I used as a training input and then on the outside on both sides this is what the flux model has generated as you can see here the details are stunning it is exactly my eyebrows as they are from the shape here you can have this like overhanging bone I have on both sides about my eyes and again it has the correct shape for that also the beard here that I often don't really shave off exactly as it looks on my face also the ear how it stands a little bit off
but it's pretty close to the head and even these short hairs here that I often don't really shave off really really looks like my actual Head super impressed by that now the page that is having this simplified and very exact process is open art and it's super easy to use you go on the website on the left you can see here models and then here you have train my own model you click on that and this is going to bring you to this page now here you can select between the style the character face and
object of course in this case we want to train a face so we click on that you want to give the name of the model and then upload images it says here between four and 500 you don't really need 500 images you're good with between four and eight images and then already you can click here on train your model just takes a little bit and then the model is already done this is how easy that is after that on the right side you have here the different models you have trained and they are private you
can also publish them if you choose to if you find something that you want to share with the world and of course another thing is you can also download it this is a function I really fought for because I I think that is super important now to download the models you have to Mouse over the image and here you have these three little dots click on them and then here you have download model and then you can just use it in any kind of local environment now when you click on download it's giving you this
kind of warning here this is just because open art a has a lot of casual users who don't use any kind of local environment like comi or fori so they might be confused and think that the model is a software or something like that so you can ignore that warning just download the model and use it like a normal Laura with flux let me also quickly show you the images I used for the training now surprisingly there is not many images of me even though I'm constantly in front of a camera I don't do a
lot of selfies so these are the kind of images and one thing I would suggest to you is to don't have extreme face expressions because this can irritate the model training and make your face look kind of strange so more timid face expressions gives a clearer better face that looks more like you now the question is how do you get more artistic style into the Laura the way I did that is by throwing in some artistic images of yourself now the way you can do that is that you go in open art and you create
with an sdxl model and a phase reference images of you in different styles for example here this looks pretty good you can actually use it like that if you want to the only thing is SXL has way less details and because of that it also looks less detail so I have created this image here I have also created this image here more in kind of like a pixer style or that image here that has more of this kind of anime style now here I want to point out something that is really important you can see
in this image that the person has a really muscular body and this can be a problem because when you use that for the flux training you will end up afterwards with images of yourself with a very muscular body because like I said flux training is very sticky to the input images so try to avoid that so that you actually get your real body shape so rather create a lot of other images with different prompt variations that suppress this really muscular body so to have more artistic style in your flux Laura simply throw in a couple
of images not too many I would say if you have five photos of yourself throw in two or three style images like paintings drawing graffiti this kind of thing that looks like you to also give a training but make sure it actually looks close to how you you look now to use the model it's pretty easy you just click here on Create and this is bringing you to the create page now here on the left side you can set all of the settings you want to do so for example you have here the model weight
which is selected for you at 0.8 but you can play around with that below that we have the prompt and when we go a little bit deeper for example here we have image guidance this is one of the tricks that I'm using I'm going to show you in a second what that does and then below that you also have the output size here and of course the create button with the amount of images you want to create for that so let's talk about the reference image here and as you can see I have this cute
character with the dangar on and I use that as an imageo image reference in this case I do have my flux Laura with the artistic style so actually wouldn't need that the reason why I'm using that is because I want to have this exact composition with these exact colors for the output image but in my own style as an image that looks like me I have a quick description here 3D cute render of a bald bearded men in dungar and when we go back down here you can see for the creativity level I have 0.9
which means I give a lot of freedom to the flux Laura to make the style of of my face and the pixa style I want to have and as you can see this is the output it looks fantastic and one thing to point out here is that the beard looks like a 3D beard not like a photo beard this is an important detail the reason why this is an important detail is this image here because when you look here it looks a lot more realistic from the fabric of the dunger but also from the beard
here which is more photographic and less like a pixel style 3D model now the way I did this here is that I used the flux model that I trained only on photos and here when we go down to the image to image reference you can see that the creativity level in this case is pretty similar so I give a lot of freedom to it but the reason why you actually need that is because if you let less creativity for the model the image is not going to look like you and there you have that big
problem that either you can have something that looks realistic like the beard hair or you can have something that looks like the pixel style but doesn't really look like you and this is why I prefer to throw in a little bit of these stylistic images into the model training to open it up and give it more freedom now let's talk a little bit about the rest of the user interface because open art has a lot of functionality that you can use to get more out of your images first of all on the right side you
can see we have appscale when you click on this you have a pop- down menu with different options in here now one thing I want to point out here and this is going to be fixed in the future I'm just saying it here because of how it works right now is when you have this upscaling plus the phas it will improve the phas detail however if the face is too big in the image it can give you strange double eyes so maybe avoid that and use just the 4X upscale without the phase Improvement it also
gives you really good quality other functions you have here are listed on the top for example you have in painting you can remove objects you can expand the image to the sides you can have here stylization so basically a style transfer you can have a background you can have a background change which is pretty cool because it basically removes the background you have right now and puts another background behind that so if you like the image of you but you think the background should look a little bit different this is the way to go and
then we have also this function here for face expressions which is a little bit hit and miss I wasn't 100% uh convinced by that but it's a kind of interesting thing to play around with now one very important thing to point out here is if you use any of these functions don't upscale the image beforehand because the image that results from these techniques is going to be a lower resolution image not the upscaled image so first do the in painting or the expansion of the image and then upscale it afterwards by the way one more
thing down here you also have the ultimate upscale and that gives you a very nice quality however it is not ideal for photos it is better for digital paintings overall I was super impressed by the quality of the luras and I'm really happy I could fight out the functions I wanted to have for you on that page and this of course is especially useful for people who don't have the hardware to run flux or train flux lauras let me know in the comments what you think leave a like if you enjoyed this video and see
you soon bye n [Music]
Related Videos
OpenAI Unveils o3! AGI ACHIEVED!
26:24
OpenAI Unveils o3! AGI ACHIEVED!
Matthew Berman
192,179 views
ULTIMATE Character Consistency Hits AI Video! (Full-Process)
19:09
ULTIMATE Character Consistency Hits AI Vid...
AI Samson
22,609 views
Google’s Quantum Chip: Did We Just Tap Into Parallel Universes?
9:34
Google’s Quantum Chip: Did We Just Tap Int...
NASASpaceNews
575,831 views
This is the Biggest AI Film Update of the Year!
47:16
This is the Biggest AI Film Update of the ...
Curious Refuge
21,052 views
How to Train a Flux LoRA Locally in 15 Minutes! Create an AI Influencer with ComfyUI Workflows.
12:16
How to Train a Flux LoRA Locally in 15 Min...
ComfyUI Studio
5,442 views
Elon Musk: "Our New Flying Cars Will Hit The Market TODAY"
9:45
Elon Musk: "Our New Flying Cars Will Hit T...
Cosmic Nexus
276 views
Get READY for the BEST AI News before 2025 - Top FREE Tool Access right now!
14:03
Get READY for the BEST AI News before 2025...
Olivio Sarikas
8,507 views
FLUX + LORA Full Course (Uncensored Super Photorealistic Images With Your Own Face)
19:42
FLUX + LORA Full Course (Uncensored Super ...
CyberJungle
330,038 views
AI Just Unlocked a NEW WAY to Colour Grade (Mind Blown!)
9:31
AI Just Unlocked a NEW WAY to Colour Grade...
Idustrial Revolution
9,707 views
This is HUGE: Video 2 Video Face Animation - Runway Act One
5:07
This is HUGE: Video 2 Video Face Animation...
Olivio Sarikas
11,278 views
AI News: The AI SHOWDOWN We Needed!
35:49
AI News: The AI SHOWDOWN We Needed!
Matt Wolfe
62,582 views
BREAKING: OpenAI's new O3 model changes everything
12:11
BREAKING: OpenAI's new O3 model changes ev...
Theo - t3․gg
136,344 views
WOW! These new AI Tools are amazing!
13:06
WOW! These new AI Tools are amazing!
Olivio Sarikas
24,805 views
Persistence is Required For Landscape Photography
14:08
Persistence is Required For Landscape Phot...
Thomas Heaton
42,194 views
GoodBye Kling AI & Minimax !! This is FREE, UNLIMITED AND BETTER
8:43
GoodBye Kling AI & Minimax !! This is FREE...
Brain Project
8,312 views
Veo 2 Hands On Review - The New King Of AI Video?
13:44
Veo 2 Hands On Review - The New King Of AI...
Theoretically Media
36,057 views
Акунин — о будущем России, праве россиян на свободу и ошибки, Путине и литературе во время войны
1:21:59
Акунин — о будущем России, праве россиян н...
Телеканал Дождь
154,731 views
OpenAI Just Revealed They ACHIEVED AGI (OpenAI o3 Explained)
12:05
OpenAI Just Revealed They ACHIEVED AGI (Op...
TheAIGRID
252,192 views
All-New 2025 Tesla Model Z: A Bold Leap into the Future
3:20
All-New 2025 Tesla Model Z: A Bold Leap in...
Drive Time
243 views
Controlling Blender with my voice using LLM
15:33
Controlling Blender with my voice using LLM
Polyfjord Deep Dive
47,818 views
Copyright © 2024. Made with ♥ in London by YTScribe.com