馃馃巻NEW YEAR EDITION馃巻馃
I needed to reupload the version due to generator issues with uploading the fixed checkpoint that I updated. Same version that has been up for the past few days, so no need to re-download if you have it already. Just added some new pictures for fun.
I've been working on this model for the past couple of weeks and felt like I got it to a point where I'd be happy to release it. I've tried to create a model that has the colors and creativity of 1.5 FaeTastic and also tried to fix some issues that I feel SDXL has. I can't promise that it's perfect, but I feel like it's a step in the right direction.
I've trained several LoRAs that haven't been released for the purpose of merging them into the model, and I'll talk about the model genetics below. Hopefully openly talking about the way I merge and create models/LoRAs will open the door for others to learn as well. :)
Because I believe in the open-source spirit of SD, anyone is free to use this model or any of my LoRAs to merge or use as they see fit in an open-source sense, including selling any images or using them in personal projects. I would be extremely grateful to be credited for any merges or uses of my resources.
However, I am very opposed to having my model and resources I personally train and test and spend hundreds of hours on, be put up on third-party generation sites for their monetary gain where I receive nothing in return.
Model Genetics
Everything was merged with SuperMerger in A1111. All LoRAs were trained with Kohya.
Dreamshaper XL .5 with Rundiffusion XL .5
Added on top of that is the Fae Style SDXL LoRA
FaeTastic V1 SDXL .7 with ProtoVisionXL .3
Added SDXL High Details LoRA
Added SDXL Better Eyes LoRA
Trained a FaeTastic SDXL LoRA on high aesthetic, highly detailed, high resolution 1.5 personal generated images and merged in. This LoRA is pretty awesome, and I'll probably release it on its own as well.
Trained an SDXL LoRA on high-resolution, highly detailed images of realistic, drawn, and digital art of hands. These images were up close pictures of hands as well as images of people holding objects. This was merged in.
Recommended Resources
For those interested in training SDXL LoRAs I highly recommend this guide that Ally and I collaborated on together with a lot of extensive testing and training: https://civitai.notion.site/SDXL-1-0-Training-Overview-4fb03d5b4cbf482ebe13ffef7bcb8ead
If you ever have any questions or need help with my resources or anything Civitai related, I am always happy to assist, you can find me in the Civitai discord! <3
Description
FAQ
Comments (33)
From version 2 to version 24? That's quite a leap! 馃構
24 for 2024! 馃槀
I have tried many other models, but this is my favorite by far.
You train with prodigy ? On Kohyass ?
With this optimizer arguments : (decouple=True,weight_decay=0.5,betas=(0.9, 0.99),use_bias_correction=False) ?
I got errors...
too many values to unpack (expected 2)
Hey there, yes when I train LoRAs. When I get a chance I'll copy and paste the arguments for you. But it's 100% prodigy
@Faeia聽Thx for the answer! I got it to run, but sometimes its horrible....
How do you prepare your training settings ? Sharpining or something like that ? Because with MJ i got really bad results. So sad
@Faeia One more question ^^ I saw you use most of the time 12/12 Epoch or most of the time the last one.
But why ? Most of the time my samples at the beginning looks very promissing, but they are bad... :D
Never tried to take the last one, because i am scared of overtraining..
@LDWorksDervlex聽When I test things, I use the version that I like the most. I don't always end up using the last epoch. It's personal preference. Just use whatever you think is best when you train and test.
I had an issue with v20 and above in comparison with v10 - it started to make seams (and it's not Upscaler seams) during upscaling. Here are some examples:
https://ibb.co/zfqLDNR
https://ibb.co/b2xY4gW
It's realy a pity, because v20 gives much more spectacular result. Is there any way to get rid of them ? An embedding or negative prompt? Or by any chance you know the cause of those and can fix it in future versions?
That's strange, it looks like a weird tiling issue that you sometimes get when upscaling. What prompts are you using and are you using any LoRAs etc? Hard to diagnose without knowing all parameters, etc. Also if you could message me on discord, (Faeia in Civitai discord) it'd be easier to discuss and figure out
@Faeia聽I think I've found what caused the problem - the aspect ratio. Starting from v20, v24 I started to get the border to the bottom and to the right during generation while using non-standard aspect ratio (1000x664) same for Upscaling tiles ratio (1200x2000), When I take the aspect ratio from your sample picture those artifacts are gone. Still I need to test it in big overnight generation.
https://ibb.co/6XZc6zM
I have tried many models, and will continue to do so. This one is still my favorite. thank you!
This model, above all others, seems to be amazing at generating that old painted illustrated look you'd find on old paperbacks, computer game boxes, VHS covers, etc. Actually looks authentic without seeming too 'digital', if you know what I mean. Nice to find a truly unique checkpoint on here every once in a while. Thanks for your work on this one Faeia!
Glad you're enjoying it, thanks for the compliment :)
Can you make a lightning version please??
I'll have to look into it when I get some time, I'm really busy currently
great feature for making a good quality images without any embedding or lora.
This is my favorite model for XL. Thanks for making it ! <3
袩褉芯褋褌芯 褋褍锌械褉 屑芯写械谢褜, 褋锌邪褋懈斜芯!
I cant thank you enough for this Checkpoint <3
perfect!
v20 has been my favourite model for a long time! I'm so happy realise there's a v24! <3 Thank you for being awesome!
Super , thank you ! i use the model a lot and am very happy with the results , Good!!
I like using this checkpoint to showcase my wildcards, it's really good. Good work :)
This is a fantastic checkpoint and produces some excellent results. Thank You. 鉂わ笍馃憤
No it's not Fantastic, It's FaeTastic
@seanbrockest聽XD
can I use this for my NSFW games / comics?
A classic. One of the best, most important, and iconic checkpoints on the site.
Why this amazing model isn't working
Please participate to the auction to make this great Checkpoint available for everybody
Hi folks, this is one of the greatest model for SDXL. Could help me to bring back on site by contributing to the auction ?
Thank you all
I didn't realize at first that Checkpoint and Lora are yours 馃槄 Thanks a LOT personally for them! I was very pleased with some of the results 馃