This One-Click FLUX LoRA Creator is a Game Changer
Vložit
- čas přidán 13. 09. 2024
- 👉 Try it yourself: pixeldojo.ai?
🚀 FLUX LoRA Creator vs SDXL: The Future of AI Art is Here! 🎨
In this video, we're diving deep into the game-changing FLUX LoRA Creator and putting it head-to-head with SDXL. Get ready to have your mind blown by the next level of AI-generated art!
🔥 What you'll see:
- Detailed walkthrough of the new FLUX LoRA Creator
- Side-by-side comparison with SDXL LoRAs
- Stunning examples of hyper-realistic, personalized styles
- Tips and tricks for getting the most out of FLUX
🌟 Why FLUX LoRA Creator is revolutionary:
- Powered by Flux Dev, the world's most advanced AI image model
- Creates LoRAs with unparalleled realism and consistency
- Perfect for artists, designers, and brands seeking unique styles
👉 Try it yourself: pixeldojo.ai/f...
💡 Have you tried the FLUX LoRA Creator yet? Share your experiences in the comments!
Don't forget to like, subscribe, and hit the notification bell to stay updated with our latest AI art tutorials and comparisons!
#FLUXLoRA #AIArt #PixelDojo #SDXLComparison #AIImageGeneration
Check out our other AI art tools and tutorials
👉 Try it yourself: pixeldojo.ai?
On the front page in pixeldojo after logging in, it says "Supports diverse aspect ratios and resolutions up to 2.0 megapixels" in the "Introducing Flux" section. That doesn't seem like much resolution - am I reading this wrong? I'm thinking in terms of total pixel count which might be wrong.
As a model it has a higher base resolution than stable diffusion 3 or stable diffusion XL. Of course you can take those base images you create and upscale them 8X using the creative upscaler.
@@allyourtechai - gotcha, thanks. that makes sense
Can I download the LoRA for offline local use once trained on Pixel Dojo?
You can't do anything unless you sub at $25 month
Thanks.
Correct, as I state in the video. These Nvidia H100’s aren’t free 😂
well worth much more than 25.00 for what you get
Obviously 🧐
@@allyourtechai For those that have a daily need for this service, its a great product. I will be using it for sure when (Hopefully) my business is ready.
I have trained 2 Loras, with zero resemblance, using pixel Dojo
That would be really surprising actually given the other 500+ customers feedback. Join the discord and I’ll help out
What you did with pixel dojo is worth every single dollar. 👏👏
Thank you!
This is awesome...just tried it and getting mixed results (some pretty good, though), but I'm sure it's because my ref photos could be better, in terms of angles and lighting. Maybe do a deep dive into how to shoot and workflow for best results? BTW, your recent UI refresh looks really pro now...great job!
Thank you. I’ll do some more tutorials for sure :)
The fact that there is no way to negative prompt, and no way to include captions, is kind of a big deal. Obviously the image generation looks great but there are limitations to what this is capable of. You saw the glasses and hat and go "Whatever works!" I mean I guess, if you're trying to create something extremely specific and it continues inferring random objects and you have no way to prompt against it, and you can't include captions, I mean let's be honest people we still have a ways to go and there is still much to be desired. The tools are awesome but paying $25 to experiment with a LoRA that may or may not even get you the results you need is a bit much in my opinion. If you go to generate an image and it decides to input random objects or things you didn;'t ask for, you are shit out of luck unless you want to start inpainting.
You just tweak your positive prompt to be more specific. Flux is incredibly good with specific prompts. The first example was literally just my trigger word. Hands down the best model I have ever used. Nothing else comes close.
I also keep getting training fail. What picture format shall we use ? I am using PNG
Black Forest labs was having capacity issues. Should be ok now
Just signed up , Training my first Flux Lora... Letsss Goooo
Welcome aboard!
Great job! This platform is great please keep up your great work!
Thank you so much!
The images appear more realistic at lower resolutions. The larger the generated image, the more the artificial, plastic look becomes apparent. Unfortunately, your site offers only limited customization options, and for portraits in the 4:5 format, an image with 896x1088 pixels is generated. However, I have found that images look most realistic at lower resolutions, such as 512x712. Therefore, we need an option to set the resolution manually. Please implement this feature. Thank you.
This is awesome.
This is the best playground platform out there!!!!!
Thank you!!!
Looks great!! Have you considered an Appsumo campaign?
Greate Job Man Well - Can I use two trained human faces in a single flux image generation?
AT LASTI! THANK YOU!!! bryan i have been encountering this issue in which any civit lora url cannot be fetched. not sure if a bug or im doing somethign wrong. im copying the link as you said in a previosy video. also i know you added the posibility of uploading loras. i tried to use sdxl and flux character ones. was i right? or is it only for styles? when trying to use them i always was forced to use a url, but the lora file was already uploaded in pixel dojo.
if you can take a look, much appreciated! i didnt pass you a list of suggestiosn or things i found because you are so fast fixing stuff and now the ux is so much better than i just decided to wait,. thanks for those daily updates and the effort you are putting into this. is a good tool, multiple uses, and with freedom above all. i do hope the community grows in pixel dojo
You have to add your api key to Civitai Lora urls. They aren’t publicly accessible without it, so my servers can’t access them.
Check out the discord for more info if you need it :)
can you use multiple LoRAs when generating?
Yep, up to 20
Danke!
thank you so much :)
Server time out or ? Tryed 3 times today to train LORA. No error messge just returns to start training after 45 minutes of 0% complete ;/ Tryed smaller images, tryed cropping too. Sucseded yesterday, but model was horizontal squezed out of proportions with my Lora of mee. Made one of a friend ( with hes permisson) and turned out ok, dont understand why mine was so squeezed and hes not. Made both the same way.
You really need to join the discord haha.
One thing I didn't see mentioned in the video... are you able to download the safetensor for use locally, or is it only usable through your service?
Yes, downloadable
@@allyourtechai You definitely did go over that on a re-watch, I'm an idiot! Thanks for the response anyways
This is great!
Thank you!
from the other side, is it possible to get consistent characters full body shot? trying it at always gives me from waist up, any advice?
With a LoRA you can typically get full body consistent characters.
@@allyourtechai i thought rather within your consistent character generations an option like this one.
Anyway, you have any place to report bug, for your information?
I need "flux realism lora". I see on your site only "flux realism" and "flux lora"?
You can add the flux realism lora by going to "my custom styles" and adding the link to it. Name it, upload a thumbnail, and you are all set.
Damn I'm amazed, long live Flux and Pixel Dojo.
Thank you!
Amazing, but 25 is too much for me. In my country is 137,00 BRL 😔
I can pay only US$ 10,00
That wouldn’t even pay for the model training
I don't know what he's smoking, $10 is more than what's needed for model training. A few hours on a high end consumer-grade GPU can do it and you can rent something much better for much less. With a service like this, you're paying for convenience.
EDIT: Looks like $25 a month, not $25 to train a model. Big difference there. $25 to do as much as you want there does seem like a deal
@@allyourtechai I agree. The service is not expensive, it is priced correctly. I am just too poor to buy it. My country charges 33% tax on supermarket consumption.
@@allyourtechai Maybe you should charge for international price parity. I believe that this way you can charge an equivalent amount for a poorer country and compensate the amount for the quantity contracted. Take Steam as an example. 😉
Hey mate, how long does it take for the image to be generated on your website for flux pro?
A few seconds using Flux Dev and Flux Pro WITH a LoRA. It's so fast that a customer today asked if it was using Schnell. My image generations run on Nvidia A100's and H100's. Both 80GB GPU's.
@allyourtechai
Great stuff. I'll subscribe soon
Isn't it free? Because even though I became a member, it was written
Subscribe To Train ... It is sad
These GPU's are $25,000 each. Why would this be free?
Also, I mention multiple times in the video the price and the subscription.
@@allyourtechai Yes, it is true. I said it might work like the free Kohya GUI Repo
ciao bryan!
bro i keep getting training fail and i am not sure what happen, I did follow the instructions.....
Check the discord. Black Forest labs is getting slammed. Same thing happened when flux pro came out. They are working on scaling for us
Thank bro it is working now🎉
Having problems with my LORA, maybe the pictures was all 1/2 quality in atificial light, hmm, the it looks exactly like me the generated images , but its squiced horizontal as yours in the video was squiced vertical with sdlx model, just a lot worse ;/ , using Flux lora so why is it so out of proportions ? maybe its better 1:1 and id been told B/W images are better for AI to interpritate ? Ill try taking better pictures when the daylight hits Denmark ;) See You ;)
make sure you are in the discord!
@@allyourtechai All cool now, made super LORA of me ;)
Can I download a LoRA for my local environment without having to upload the LoRA I created to HugginFace ?🤔
On pixeldojo nothing ever goes to hugging face. Download straight from pixeldojo. You can then delete the file from my servers by hitting the delete button, or you can use it to generate images on the site
@@allyourtechai Thank you~~☺
I cant zip it. Which program do you use?
i recieved model training failed
I just use the zip utility built into windows
If it's the one I see in the logs you need to use regular .png or .jpg images. Join the discord. Much easier to chat there.
I keep getting upload failed
Some context would be helpful
Ah fuk it i think when i sub hime thet is normal ytuber...but now i see that he adver just his pixle jojo shit....have fun here
Been waiting for this feature, I had a feeling you where onto this ;) Nice now wee dont have to go Hugging Face or Civit and put ouer personel LORA :) Thats just awsome ;) !Thank You Dr. Manhatten ;)
Haha I’ve been working away at this one for a bit now, so glad it is live!