AI illustrations of DOAXVV chara.

[Hans Jürgen]

If someone finds out how to do this with EasyDiffusion, please share ! Thanks

[exorcismus1]

39 minutes ago, 5150 said:

Hello.

This Lora is Nanami Ver.3.

I post a smaple of the output 8 Batch size.

You can see in these images, the success rate is not high.

So I think it is necessary to devise a way to make Lora.

 

Prompt

nsfw,best quality, photorealistic, masterpiece, 8k, high res, solo, 1woman, (26 years old:1.2), (looking at the viewer:1.2), (professional lighting), low angle view, perfect female figure, living room, ceiling, night,  (dynamic pose:1),stylish pose, soft lighting,
(public hair:1.4),(detailed face, black hair,brown eyes:1.2), (full body:1.2),middle breast,skinny,(nude:1.2),
<lora:DOAXVV_Nanami3:0.85>, <lora:jkPerfectBoobs_V02:0.15>

 

Negative prompt

Imperfect face,(bad eyes),(Bad eyelashes),Asymmetric eyes,Blurred eyes,Unclear eyelashes,abdominal muscle,big nose,paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2), lowres, ((monochrome)), (grayscale:1.2), head out of frame, acnes, (skin blemishes:1.2), age spot, glans,extra fingers,fewer fingers,(watermark:1.2),(letters:1.2),

 

Model:Chilloutmix

Sampling method:DPM++ 2M Karas

 

Maries's LORA has a higer success rate than Nanami's LORA.

But Amy's LORA has a higer success rate than Marry's LORA.

DOAXVV_Nanami3.safetensors 165.42 MB · 3 downloads

thank you i will try it out btw it's very hard +cry

[Hans Jürgen]

1 hour ago, 5150 said:

Hello.

Thanks for your message.

I also try to make AI illustrations after hearing from my friends last month.

Only input 1 word, It's change images quickly.

I found DOAXVV Chara's Lora  in Civitai. But almost Lora were too influenved by LORA.

For example pose, clothes etc.

So I try to make new Lora...

But It's difficult.

I hope that we share knowledge to use AI is good!

 

Thank you.

Man, I´d *like* to share knowledge.

But sadly I could not reproduce any of your glorious results. I could not make any DOAXVV face at all... with two different SD versions installed on my PC...

Guess I am too stupid (for now)

Keep going, 5150

[himawari56]

Thanks to fiona Lora for sharing by 5150, I want to train Sayuri's Lora when I have time. but it may be a while.

[5150]

16 hours ago, himawari56 said:

Thanks to fiona Lora for sharing by 5150, I want to train Sayuri's Lora when I have time. but it may be a while.

Hello. You are welcome.

Very beautiful pics!

 

Thank you.

 

Edited April 9, 2023 by 5150

[5150]

[5150]

[5150]

 

 

Edited June 9, 2023 by 5150

[5150]

Edited April 14, 2023 by 5150

[Zantsetsuken]

Which words with stable diffusion are you using to genetate thoses images??? i just got horrific pictures directly from nightmares when typing : marie from DOAXVV

[zakunhope]

Tamaki  ?

[5150]

2 hours ago, zakunhope said:

Tamaki  ?

very beautifull pisc!

[lamialove]

[5150]

1 hour ago, lamialove said:

Yukino!! Very similar eyes!

[geniuseast]

Hi, I am training and testing Lora as well. Have tried 2o epochs with or without regulation images. Here are some results with Nanami. Let me know how you feel.

[Hans Jürgen]

I was still not able to recreate a single DOAXVV face... damn

Would you nonetheless consider releasing your Lora of Nagisa, 5150 ??

Anyway, please continue your good work, 5150 !!

[5150]

18 hours ago, geniuseast said:

Hi, I am training and testing Lora as well. Have tried 2o epochs with or without regulation images. Here are some results with Nanami. Let me know how you feel.

Hello. How many ScreenShot did you use as Bese pic?

In order to improbe reproducibikity, I am also doing trial and error to determine how many photos to use as a base,How many epochs,how many times to learn each photo, and what kind of photos are good...

Edited April 19, 2023 by 5150

[Hans Jürgen]

6 hours ago, 5150 said:

Spoiler

Hello. I understood how to make Lora from these informaitons.

 

1

https://dolls.tokyo/about-kohyas_gui-setup/

There is some old information about install steps in this page,

 

2

so I used this information to install.

(https://github.com/bmaltais/kohya_ss)

 

3

But it is not enough information how to make Lora above pages. So I used this video as a reference.

This videos is especially useful for how to use Tigger and how to edit text file.

 

DOAXVV_Nagisa_Pony（0.85）

 

DOAXVV_Nagisa_Short（0.85）

 

DOAXVV_Nagisa1（0.9）

DOAXVV_Nagisa_Pony.safetensors 165.42 MB · 2 downloads DOAXVV_Nagisa_Short.safetensors 165.42 MB · 2 downloads DOAXVV_Nagisa1.safetensors 10.6 MB · 3 downloads

 

 

Thanks for your answer, and especially thank you very much for your Nagisa tensors!!!!

I have a working version of Automatic Stable Diffusion, and can generate lots of cute pictures. But never do they look like Marie or Amy or Nagisa...

Maybe a keyphrase is missing, a trigger ?

Anyway, I will try again with my waifu Nagisa and your Loras, thanks again for sharing, 5150! You´re one of the good guys!!

[geniuseast]

6 hours ago, 5150 said:

Hello. How many ScreenShot did you use as Bese pic?

In order to improbe reproducibikity, I am also doing trial and error to determine how many photos to use as a base,How many epochs,how many times to learn each photo, and what kind of photos are good...

 

Hi, I used 20 Screenshots as base pics. I found that the results depend on how you crop the photo, if you did not put the character in the middle it will end up like pic 3 and 5 above. And you cannot fill the image with (a solid color e.g black) when you crop as the black could appear when you produce images.

 

I also tried using some regulation images, pic 4 and 5 were produced with lora training by regulation image (the class is person). It was training by 40 epochs. I feel like not too much difference comparing to 20 epochs with no regulation.

 

One thing I really struggle with is how to let the doa character from the game looks more realistic. I feel like you did better than me.

 

[zakunhope]

Nanami 2.5D style

[linrjun]

it's really difficult to get started with AI, especially whiout a best spec PC.

I found the flaws on hands unsolvable.

[zakunhope]

[fenix71]

On 4/19/2023 at 2:12 PM, 5150 said:

Hello.

Don't need keyphrase. for example

 

Prompt

nsfw,masterpiece, ultra high res, (photorealistic:1.6), 8k, RAW photo, best quality, super detail, detailed shadow, realistic and soft lighting, lens glow, depth of field, sharp focus,
1girl, (small breasts:1),
(temple),(white gothic dress:1.1),
(looking at viewer,full body:1.2),
light smile,  <lora:DOAXVV_Nagisa_Pony:0.85>

 

NP

delight, bright, painting, sketches,  (portrait: 1.6), ((bad legs,more than two legs,bad anatomy:1.5)),(worst quality:2),(low quality:2),(normal quality:2),lowers,normal quality,((monochrome)),((grayscale)),skin spots,acnes,skin blemishes,age spot,ugly face, fat,missing fingers, extra fingers, extra arms, delight, bright, makeup, facemask,(mosaic censoring:1.4), veil, hair hair ornament,

 

only input red word.

 

 

Prompt

nsfw,masterpiece, ultra high res, (photorealistic:1.6), 8k, RAW photo, best quality, super detail, detailed shadow, realistic and soft lighting, lens glow, depth of field, sharp focus,
1girl, (small breasts:1),
(temple),(white gothic dress:1.1),
(looking at viewer,full body:1.2),
light smile,  <lora:DOAXVV_Marry:0.85>

 

NP

delight, bright, painting, sketches,  (portrait: 1.6), ((bad legs,more than two legs,bad anatomy:1.5)),(worst quality:2),(low quality:2),(normal quality:2),lowers,normal quality,((monochrome)),((grayscale)),skin spots,acnes,skin blemishes,age spot,ugly face, fat,missing fingers, extra fingers, extra arms, delight, bright, makeup, facemask,(mosaic censoring:1.4), veil, hair hair ornament,

 

 

Prompt

nsfw,masterpiece, ultra high res, (photorealistic:1.6), 8k, RAW photo, best quality, super detail, detailed shadow, realistic and soft lighting, lens glow, depth of field, sharp focus,
1girl, (small breasts:1),
(temple),(white gothic dress:1.1),
(looking at viewer,full body:1.2),
light smile,  <lora:DOAXVV_Amy:0.83>

 

NP

delight, bright, painting, sketches,  (portrait: 1.6), ((bad legs,more than two legs,bad anatomy:1.5)),(worst quality:2),(low quality:2),(normal quality:2),lowers,normal quality,((monochrome)),((grayscale)),skin spots,acnes,skin blemishes,age spot,ugly face, fat,missing fingers, extra fingers, extra arms, delight, bright, makeup, facemask,(mosaic censoring:1.4), veil, hair hair ornament,

 

 

Only change Lora file name.

Thank you.

 

DOAXVV_Nagisa_Pony.safetensors 165.42 MB · 56 downloads

 

Please which tool do you use to enter this prompts ?

[5150]

 

 

 

Edited June 8, 2023 by 5150

[IRqXzFmB0]

22 hours ago, 5150 said:

 

 

 

Why do you erase every single thing?