sorry if this is OT, but i'm really impressed by these AI fan art
how do you do this? is there some kind of a guide to learn how to do it?
how do you do this? is there some kind of a guide to learn how to do it?
- Dec 15, 2020
- 2,163
- 4,921
Try putting the word "guide" into the search bar up in the corner and see what comes up searching through the thread.
- Jul 3, 2021
- 4,565
- 22,294
This post contains the link to the website where you can create your own...
https://f95zone.to/threads/our-red-string-fan-art.111327/post-7753847
- Dec 15, 2020
- 2,163
- 4,921
Surprising, given that I find a few good comments that can get you started. When I played around with the tools that's how I got started.
Tried Anime style today, what do you guys think ?
- May 28, 2019
- 930
- 2,222
Well, it's certainly interesting and better than realistic ones since it's still "cartoon" style (please don't kill me weebs) and doesn't break immersion from original game, but I think this style makes characters look older!?
Realistic renders based on the character cards.
- Jan 14, 2018
- 212
- 819
- Jul 3, 2021
- 4,565
- 22,294
- May 11, 2020
- 74
- 292
You must be registered to see the links
and if you got an off the shelf Walmart computer like I do, here's a free and fast method but holy heck, do you run through the free daily credits fast.
You must be registered to see the links
So far, I've spent weeks trying to bring Ashley Graham to the ORS Universe.
- Jun 10, 2019
- 256
- 5,284
Hello everyone, I have worked on Chapter One of Cindy's alternative history, I have decided to publish a Teaser, the final Chapter will be published the last week of December or second of January, I want the chapter to be about 30 minutes, I will be informing you here and for my Patreon, thank you for your support
_____________________________________
_____________________________________
You must be registered to see the links
- Jan 9, 2019
- 127
- 556
People have been generating with my Lora on civit.ai, wanted to thank you guys as seems, I get back Buzz, civitai currency from it, when you use their site to generate online. which means I have free buzz to train more lora, Thanks.
We have been using Pony a variation of SDXL. The variations to put it basically are different way models are trained, could be tagging, could be data, could be other parameters, how the data is fed. Recently there is a new variation called Illoustrious, as the name suggest and in my opinion it is much better at doing Illustration stuff.
Which means even though I didnt want to train before new update. I wanted to do a 25 epoch training, my last PONY evakiss one was 75. to do a comparison. And I am going to post the results here, I actually made a Fellatrix lora which is what impressed me and made me want to do a lora of Evakiss's style. Benefits include no more pony score when tagging which is something.
the results I like it more than my last few Pony loras as I really think I can see the improvement. I will do a real big training when there is a new update. I tried to get better result my resizing all images to 1024x1024 which did effect the training, next time I will do full 1920x1080.
I am hoping doing full resolution and higher epochs will make the faces better, you guys should already know the problem there due to dataset, I also somehow have to cut down the images to about 200 or 300 at max, redundancy and similiar files. if you want to train lora same way as you see, artstyle can be "learned" even in 10 or 20 epoch, its like a sine wave, it can learn fast but details take many cycles.
Its live, since I do think its good, use Eular A or Restart if your thing can do it, I think Restart is very good at getting good result first try with less errors.
all mine are one shot than upscaled 1.5, so these are basic generation yet looks good enough I feel. I use Fooocus as I have said before.
Make sure you are using a Illustrious checkpoint
I used
We have been using Pony a variation of SDXL. The variations to put it basically are different way models are trained, could be tagging, could be data, could be other parameters, how the data is fed. Recently there is a new variation called Illoustrious, as the name suggest and in my opinion it is much better at doing Illustration stuff.
Which means even though I didnt want to train before new update. I wanted to do a 25 epoch training, my last PONY evakiss one was 75. to do a comparison. And I am going to post the results here, I actually made a Fellatrix lora which is what impressed me and made me want to do a lora of Evakiss's style. Benefits include no more pony score when tagging which is something.
the results I like it more than my last few Pony loras as I really think I can see the improvement. I will do a real big training when there is a new update. I tried to get better result my resizing all images to 1024x1024 which did effect the training, next time I will do full 1920x1080.
I am hoping doing full resolution and higher epochs will make the faces better, you guys should already know the problem there due to dataset, I also somehow have to cut down the images to about 200 or 300 at max, redundancy and similiar files. if you want to train lora same way as you see, artstyle can be "learned" even in 10 or 20 epoch, its like a sine wave, it can learn fast but details take many cycles.
You must be registered to see the links
Prompts used for the images are there if you go there on the images.Its live, since I do think its good, use Eular A or Restart if your thing can do it, I think Restart is very good at getting good result first try with less errors.
all mine are one shot than upscaled 1.5, so these are basic generation yet looks good enough I feel. I use Fooocus as I have said before.
Make sure you are using a Illustrious checkpoint
I used
You must be registered to see the links
but any of them should work. I am new to illustrious myself. but Pony days are done I guess.