[ yn / yndd / fg / yume ] [ o / lit / media / og / ig / 2 ] [ ot / cc / x / sugg ] [ hikki / rec ] [ news / rules / faq / recent / annex / manage ] [ discord / matrix / scans / mud / minecraft / usagi ] [ sushigirl / lewd.sx / lainzine ]

/o/ - Art / Oekaki

Oekaki is back!!!
[catalog]

Name
Email
Subject
Comment
File
Password (For file deletion.)

News Post: I am Retiring.

File: 1725533322207.jpg (276.42 KB, 2048x2048, madotsuki_ponyrealism_esrg….jpg)

 No.5388

I know, IA generated content is controversial and many don't consider it to be art, but i feel depressed just by looking at this one.

Prompt is embedded in the picture if needed.

 No.5389

File: 1725535746061.png (1 MB, 1024x1024, 00058-3089465979.png)


 No.5390

AI is fun to play with. Made these Monoe chibis yesterday.

 No.5391

>>5390
3>4>1>2

 No.5392

>>5390

Didn't know there was a Monoe LoRa for Pony.

Just found out there's a Monoko one too. I wonder how horryfying it would look with Pony Realism.

 No.5393

>>5392
> I wonder how horryfying it would look with Pony Realism.
Give it a try and show us the results.

Also, two more images I made earlier and another two with an addition of some questionable artist's style LORA.

 No.5394

>>5393

Hum, doesn't seem to be compatible, guess it was developped on the SD 1.5 Pony version.

 No.5395

File: 1725570055214.png (809.77 KB, 1072x1072, 00123-3152414587.png)

But Monoe LoRa works.

Kind of cool and really cute.

 No.5396

>>5394
Try to include something about realism in the prompt maybe? This might work, I've started to mess with Pony Diffusion just a few days ago.

Currently upscaling this one.

 No.5397

>>5396

The issue isn't the prompt, A1111 doesn't even see the LoRa when i load Pony Realism (which is based on SDXL).

I guess it would work with the original Pony models based on SD 1.5 but i don't use them anymore (and it won't achieve the realistic result i'm seeking anyway).

 No.5398

File: 1725570550602.png (871.12 KB, 1072x1072, 00126-3152414587.png)

And here goes "nightmare fuel" Monoe.

 No.5399

File: 1725570684735.png (835.14 KB, 1072x1072, 00127-3152414587.png)

Now cute again.

 No.5400

File: 1725571070493-0.png (246.38 KB, 1065x1175, Untitled1115_2024082223581….png)

Weirdly enough, Pony Diffusion model works significantly slower than Waifu Diffusion that I used to use before. Like, a simple 512x768 image generation takes ~3 minutes, while with Waifu Diffusion it was taking a mere seconds on my RTX 4060.

I'm suspecting sampling method might be an issue (with WD I used Euler a, while with PD I used DPM++ 2M).

 No.5401

"depressed"?

 No.5402

>>5400

It only takes a dozen of seconds on my RTX 3090 even using hi res fix with the same sampling method. I was thinking about a lack of memory but CUDA should throw an error if it runs out if VRAM… Or maybe it falls back on the CPU ?

 No.5404

Ok, so Waifu Diffusion weigth around 3.5GB in full precision while Pony XL is around 6.5GB, this hits quite hard on a 8GB card. Try using the –medvram flag if you're running A1111, it won't do any miracle but it should help.

 No.5405

>>5402
The issue has fixed itself after I tried to play with a few other models and then returned to the one that was working slow. Weird.

 No.5407

File: 1725609699003.png (2.96 MB, 1216x2712, 00094-3089465979.png)


 No.5408

File: 1725612786754.png (659.1 KB, 1072x1616, 00031-3152414587.png)

Cute bikini Monoe.

 No.5409

File: 1725621439790.png (609.52 KB, 1024x1536, 00088-3152414587.png)


 No.5410

File: 1725621847928.png (1.12 MB, 1024x1536, 00092-3152414587.png)


 No.5411

File: 1725622930397.png (1001.4 KB, 1024x1536, 00108-3152414587.png)


 No.5412

File: 1725623509422.png (1.23 MB, 1024x1536, 00120-3152414587.png)


 No.5413

File: 1725624149129-0.png (925.42 KB, 1024x1536, 00128-752878247.png)

File: 1725624149129-1.png (1.04 MB, 1024x1536, 00130-1313273604.png)


 No.5414

Seems like this is gonna be Monoe thread for now.

 No.5415

File: 1725655471585.png (1.05 MB, 1024x1536, 00053-1313273604.png)

>>5414

Monoe is love.

 No.5418

File: 1725698077735.png (928.63 KB, 1072x1072, 00017-3516057778.png)


 No.5419

File: 1725702164998-0.jpg (339.65 KB, 3440x1440, monoe_wallpaper_1.jpg)

File: 1725702164998-1.jpg (322.99 KB, 3440x1440, monoe_wallpaper_2.jpg)

File: 1725702164998-2.jpg (267.61 KB, 3440x1440, monoe_wallpaper_3.jpg)

Some Monoe wallpapers for great justice.

 No.5420

File: 1725704606185.jpg (291.5 KB, 3440x1440, sabitsuki_wallpaper.jpg)

Sabi hijack.

 No.5425

>>5423

Mixing LoRas feels like playing mad scientist.

 No.5430

>>5428
Why does it look like…
Looks suspiciously like…

 No.5433

>>5430
That's exactly what you think it is.

 No.5434

What has science done

 No.5435

File: 1725988094483.png (1.28 MB, 1536x1024, 00075-258907493.png)


 No.5436

File: 1725988593416.png (931.19 KB, 1024x1024, 00010-1108610915.png)


 No.5437

File: 1725998525210.png (1.02 MB, 1024x1024, 00095-1515613763.png)

The more, the merrier.

(Also, CivitAI seems down/extremely laggy).

 No.5438

File: 1726002074839.png (1.03 MB, 1024x1024, 00141-1515613763.png)


 No.5448

>but i feel depressed just by looking at this one.
seriously what did OP mean by this?

 No.5449

>>5448
He wishes Mado were real

 No.5450

>>5448

Its realism and eeriness conveys the feeling of hopelessness and depression feeled by madotsuki.

 No.5451

File: 1728112286421.png (441.81 KB, 1024x768, 00006-1424909167.png)


 No.5452

File: 1728113000364.png (481.78 KB, 1024x768, 00014-1424909167.png)

Fiddling with regional prompting.

 No.5453

File: 1728212731339-0.png (1.49 MB, 1024x1536, image.png)

>>5452
I was able to make a similar image with no regional prompting.

 No.5454

File: 1728212943170-0.png (1.49 MB, 1024x1536, image.png)

Whoops, forgot to remove an extra hand.

 No.5455

>>5453

Yes, but it's tricky, uncertain when you want multiple characters and it's really hard to mix different characters LoRas.

 No.5456

File: 1728218265896.png (2.71 MB, 2048x1536, 00037-4186273829.png)


 No.5457

File: 1728219010774.png (2.88 MB, 2048x1536, 00044-4186273829.png)

I love this one.

 No.5459

File: 1728588737911.png (1.34 MB, 896x1152, 00031-114321773.png)

Introducing Flux.1.

DiT models are extremely impressives, but hit quite hard on GPU and eat truckloads of VRAM (weights of the original Flux-Dev model are around 24GB).

A111 users can switch to Forge if they want to give it a try.

 No.5460

File: 1728589453701.png (1.16 MB, 896x1152, 00033-652040992.png)

Unlike SD models, it performs incredibly well in generating readable text.

 No.5461

File: 1728590292594-0.png (1.59 MB, 1024x1536, image.png)

File: 1728590292594-1.png (1.56 MB, 1024x1536, image.png)

>>5459
>>5460
Sounds good, but Pony Diffusion serves me just fine for now.

 No.5462

>>5461

Well, SDXL based models have tons of LoRa, are versatiles and still quite capables (indeed Pony is pretty good), but diffuser-transformers are a big step forward.

 No.5463

File: 1728851126612-0.png (Spoiler Image, 568.71 KB, 1024x1536, 00012-1763326709.png)

File: 1728851126612-1.png (Spoiler Image, 536.4 KB, 1024x1536, 00011-894256726.png)

>>5462
But can it do this?

 No.5464

File: 1728921107493.png (Spoiler Image, 1011.63 KB, 896x1152, 00004-2250167651.png)

>>5463

Yes, if you want to have nightmares, it can do way worse.

 No.5465

File: 1728922738534-0.png (Spoiler Image, 1.02 MB, 896x1152, 00031-3029348520.png)

File: 1728922738534-1.png (Spoiler Image, 1.01 MB, 896x1152, 00033-518874898.png)


 No.5471

File: 1729023404373-0.png (2.57 MB, 1536x2304, 00039-1241394522.png)

File: 1729023404373-1.png (3.04 MB, 1536x2304, 00040-1495300763.png)


 No.5472

im gonna kam to these pictures

 No.5473

I'm going to train a monoe LoRa for Flux (hoping my card has enough VRAM), I'll upload it on Civit.AI when done.

 No.5474

File: 1729086105348-0.png (147 KB, 512x512, monoe-flux_000100_00_20241….png)

File: 1729086105348-1.png (203.53 KB, 512x512, monoe-flux_000900_00_20241….png)

>>5473

So far, so good.

First picture is prompt at start of training, second after ~800 steps.

 No.5475

File: 1729096533346.png (345.5 KB, 512x512, monoe-flux_000800_00_20241….png)

Still in training, starting to look pretty good.

 No.5476

File: 1729100659836-0.png (255.66 KB, 512x512, monoe-flux_002200_00_20241….png)

File: 1729100659836-1.png (174.63 KB, 512x512, monoe-flux_002800_00_20241….png)

Will be done in about an hour.

 No.5477

File: 1729101000028.png (293.67 KB, 512x512, monoe-flux_000400_00_20241….png)

Short hair Monoe looks awesome.

 No.5478

File: 1729107661894.png (810.25 KB, 896x1152, 00024-1241394522.png)

Done, but not really satisfied.

 No.5479

File: 1729108594862.png (Spoiler Image, 868.95 KB, 896x1152, 00043-3028483749.png)

A sexy one.

 No.5480

File: 1729110752207.png (Spoiler Image, 903.86 KB, 896x1152, 00050-3028483749.png)

>>5479

And finally, a ecchi one (spoiler : nipples), keep in mind that Flux wasn't really made for this.

 No.5481


 No.5482

File: 1729113373898-0.gif (1.96 MB, 498x280, ltg.gif)


 No.5483

File: 1729546888378.png (588.73 KB, 896x1152, 00034-4253932786.png)

First shot at a Monoko LoRa for Flux Dev.

Will run more tests tomorrow and maybe upload it to civitai.

 No.5484

File: 1729547539509.png (746.7 KB, 896x1152, 00041-3140741274.png)


 No.5485

File: 1729587790488.png (699.75 KB, 896x1152, 00000-2357520426.png)


 No.5486

File: 1729588421612.png (1.02 MB, 896x1152, 00008-1164126834.png)

Here goes the "toilet benchmark".

I think i'll retrain it on the best generated pictures.

 No.5487

File: 1729604800174.png (Spoiler Image, 936.69 KB, 896x1152, 00067-1529143786.png)

Can't call this "sexy" nor "ecchi" by any means, but here's nude Monoko.

If think the LoRa is accurately catching the concept behind this character.

 No.5488

File: 1729625650996.png (Spoiler Image, 970.02 KB, 896x1152, 00100-1529143786.png)

>>5487

And here's the V2, same prompt as the previous nude monoko.

 No.5489

File: 1729627455561.png (649.41 KB, 896x1152, 00109-1674407883.png)


 No.5495

File: 1729764840018-0.png (1.85 MB, 896x1152, 00007-4020079655.png)

File: 1729764840018-1.png (1.69 MB, 896x1152, 00020-4237170242.png)

Voynich manuscript lost pages, depicting Monoe and Monoko.

 No.5537

File: 1730934994785-0.png (2 MB, 1152x896, 00003-2872367896.png)

File: 1730934994785-1.png (1.16 MB, 1152x896, 00006-2314966260.png)

File: 1730934994785-2.png (2.21 MB, 2048x896, 00009-337895557.png)

So this isn't YN related in any way, but here's a low rank adapter of one of the creepiest boss in video games history.

I'm amazed at how well it is able to generate some seriously scary shit despite having been trained on such a small dataset, Flux is really something.

 No.5545

File: 1732915415223.gif (1.72 MB, 480x720, monoko.gif)


 No.5546

File: 1732917311959.gif (1.91 MB, 480x720, monoe.gif)


 No.5548

File: 1733002858115-0.png (493.93 KB, 1024x1024, 00001-287542328.png)

It's been a while.

 No.5549

File: 1733087929447.gif (1.72 MB, 349x478, omnom.gif)

>>5548

Have been fiddling a lot with ComfyUI nodes.

 No.5551

File: 1734307519904-0.png (901.51 KB, 1024x1024, 00023-1014568263.png)

File: 1734307519904-1.png (1.18 MB, 1024x1400, 00037-533382233.png)

File: 1734307519904-2.png (981.32 KB, 1024x1400, 00042-2874356957.png)

File: 1734307519904-3.png (758.43 KB, 1024x1024, 00065-17632260.png)

>>5549
I've tried to mess with ComfyUI, but quickly gave up on it and returned back to good old Stable Diffusion. One good fella suggested to use SD Reforge and perturbed attention guidance plugin, have been using them ever since and I'm pretty happy with the results.

 No.5552

>>5551
Jesus christ you should all be shot. Go back to twitter if you want to soy out over your faggy algorithms(USER WAS FED TO AI MODEL FOR THIS POST)

 No.5561

>>5551

ComfyUI is harder to master, but its shit tons of available nodes and the control it gave you over the models is unparalelled.

SD webui is perfectly fine for generating "simple" pictures, but once you step into complex productions, this is another story.

There's a lot of workflows available on the net you can use as bases.

Anyway, happy new year, hope this year we'll get even more tensor power and VRAM.

 No.5563

File: 1737803389726.jpg (91.82 KB, 581x477, 1737635517773.jpg)


 No.5568

File: 1741858338209.png (147.16 KB, 488x680, ClipboardImage.png)

>>5388
you are all failures. every movement you make is an insult to the human spirit. you shall all be smote by the fires of the kami

 No.5569

File: 1741993294393-0.png (1.42 MB, 1024x1536, 1655300443857.png)


 No.5573

File: 1744780430624-0.png (1.33 MB, 1024x1536, 1618296884901.png)

File: 1744780430624-1.png (1.52 MB, 1024x1536, 1612363757669.png)

File: 1744780430624-2.png (1.45 MB, 1024x1536, 1656774249403.png)

File: 1744780430624-3.png (1.53 MB, 1024x1536, 1731648758852.png)

I have discovered Illustrious model and oh my goodness gracious, it's so much better than PD.

 No.5577

>>5573

It's another model based on SDXL and yes, Illustrious is quickly taking over Pony, a shit ton of LoRa is already available for it.

Personally, I've switched to Flux, Diffusion Transformers are way more powerful than U-net based models.

 No.5578

File: 1746697375376.png (1.52 MB, 1024x1536, 00426-2972012674.png)

>>5577
Not like you need a LoRa anyway, at least for art styles. HassakuXL (the one I use) can imitate art styles of pretty much any somewhat popular artist.
> Flux
It still can't do anime style, isn't it?

 No.5579

>>5578

I was more referring to characters LoRa.

And Flux can do anime style image/characters without any problem, I've started training a few LoRas for specific characters (like the Monoe and Monoko ones I've posted earlier).

There's also some specific checkpoints of F1 dev that have been trained to produce only anime/manga pictures, pretty much like Pony and Illustrious.

 No.5580

File: 1746707193622-0.png (1.7 MB, 1024x1536, 1705030451704.png)

>>5579
I've had no need of characters LoRa either, as the model I use recognizes all the (somewhat) popular characters too.

Good to know Flux can do anime style, but is it still requires more than 8 gigabytes of VRAM? If that's the case - guess I'll stick with Illustrious for now.

 No.5581

>>5580

Q4 quantized version of Flux Dev only needs around ~6.8GB of VRAM to run and produces pretty good results.

I've tested it on a mobile RTX 4060.



[Return][Go to top] Catalog [Post a Reply]
Delete Post [ ]
[ yn / yndd / fg / yume ] [ o / lit / media / og / ig / 2 ] [ ot / cc / x / sugg ] [ hikki / rec ] [ news / rules / faq / recent / annex / manage ] [ discord / matrix / scans / mud / minecraft / usagi ] [ sushigirl / lewd.sx / lainzine ]