[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


Previous thread: >>101104870 (Cross-thread)

>What is DALL-E 3?
It's a text-to-image generator made by OpenAI.

>Links
https://www.bing.com/images/create
https://designer.microsoft.com/image-creator
https://copilot.microsoft.com/

Bing AI Slop (0.7.0)
https://pastebin.com/raw/0rre6VvS (embed)

Designer AI Slop (0.5.1)
https://pastebin.com/raw/eUSVC7Ew (embed)

>How does it work on Bing?
In Bing's Image Creator, "boosts" are used to speed up the image creation process, each user starts with 15 boosts. When you use a boost, the AI creates your image quickly. If you run out of boosts, the image waiting time increases, ranging from seconds to minutes.

>Resetting Boosts
Try clearing your search history at https://www.bing.com/profile/history

>Landscape and Portrait regeneration
Follow the instructions here: https://rentry.org/bicedit

>Prompt Generator
https://tipseason.com/dalle-prompt-generator

https://files.catbox.moe/o4yihf.png
https://files.catbox.moe/320rm1.png
https://files.catbox.moe/8gge6o.png
https://files.catbox.moe/dcokjf.png

>Related Boards
>>>/v/dall-e
>>>/co/dall-e
>>>/trash/aibf
>>>/aco/dall
>>>/tg/slop
>>
>>101143469
It works, but I can already tell the landscape/portrait gens are still more strict with filters than 1:1. I've got prompts that spit out 3-4 images all day with 1:1 that never manage to gen anything with portrait/landscape. Too bad.
>>
>>
>>
>>
>>101146627
he cute
>>
File: bing240624-141857_3572.jpg (150 KB, 1024x1024)
150 KB
150 KB JPG
>>
>>
Widescreen designer tests with old prompts, lol.

https://files.catbox.moe/6yvyzc.jpg
https://files.catbox.moe/mcibpl.jpg
https://files.catbox.moe/dy9di6.jpg
>>
>>
>>
>>101146697
I don't prompt a lot of stuff that would get dogged, but the few that I do seem to work better with this new Designer feature than when regenerated, at any rate.
>>
>>101148728
Some of them work better than others.
>>
>>
>>101148336
Hey man I swear I didn't know that cyberweed had a backdoor
>>
>>101148336
>>101149561
Start using widescreen, your movie shots will look a lot more authentic. Movies aren't square.
>>
>>101149608
the problem with widescreen is that d3 like most current LLM models, were trained on images that were preprocessed to 256x256 resolution during training.
so, the upscale to 1024x1024 is strenuous enough but when you start adding in a format that is not square but a rectangle, you really start to notice the issues. 1024x1792 or 1792x1024 really punches the most out of the model and I personally don't think the results looks good.
>>
>>101149608
Don't worry I landscape nearly every pic, but half the time it turns out worse than the original, since it's just another layer of RNG
>>
>>101149674
I think they look alright from a technical standpoint, they're just kinda annoying. You try doing a pic that looks neat but then the wide version is either just lamer or has weird issues. Or vice versa. Basically rolling the dice twice
>>
>>101149674
I've observed issues with portrait (1024x1792) format, long necks and other weird aberrations are common enough, both when "regenerating" the picture and when prompting directly in portrait mode, making it an exercise in frustration; but I can't honestly see a lot of issues with landscape mode.
>>101149833
>>101149901
Prompting *directly* in landscape mode with Designer would remove that additional layer of RNG and it would give you up to 4 results per attempt, which is much less frustrating than regenerating one picture at the time.
>>
File: 2024-06-25-2.jpg (248 KB, 1024x1024)
248 KB
248 KB JPG
Been watching some cool arty videos with acrylics and blending the colours and such.
It's alright, but needs working on the prompt (which I just remembered I never saved, ffs).
>>
>>101149901
Like here's one that actually managed to stay fairly similar across all versions, but then I would probably still say I prefer the square one. It's cool that the wide one is wide but it's not as good imo
>>
File: OIG1.jpg (453 KB, 1792x1024)
453 KB
453 KB JPG
>Monotype printing, manga art by Yoshitaka Amano, gouache, vivid colors. Comely young French woman (chestnut hair with one braid, blue eyes, white porcelain soft skin complexion) trekking in an idyllic calanque in a sunny day, bright blue sky, summer vibes. Her wolf following her.
>monotype print

The rendering is nice but i think we can use this style in a much more interesting way.
>>
>>
>>
>>
>>
>>
File: 7c13.jpg (172 KB, 1024x1024)
172 KB
172 KB JPG
>>
File: document.png (2.3 MB, 1024x1024)
2.3 MB
2.3 MB PNG
is its capability still fucked for anything even remotely suggestive? i used to be able to get pics like this out of it, just cute curvy girls in bodysuits, then the swiftpocalypse nuked all my prompts
>>
>>101151518
kek
>>
>>101151871
Go use Cici and you'll have no problems getting bodysuits and such.
>>
File: 1694611437452571.png (1.32 MB, 1024x1024)
1.32 MB
1.32 MB PNG
>>
>>101152621
cici says it's banned in the us?
>>
>>101152786
Hop on a VPN server located in Asia and you'll be good to go.
>>
>>101152902
where in asia? free vpns are rather limited in their exit node countries. i'm tryin windscribe since they have a free hong kong servier but that doesn't seem to be one of the approved locales
>>
>>101153524
I use Opera's built-in one on the "Asia" server. It doesn't mention a country so I'm not sure what locale.
>>
>>101153869
guess ill give opera a download
is this opera gx or classic opera?
>>
File: midcuck.png (275 KB, 426x464)
275 KB
275 KB PNG
is dall-e any better than this shit
>>
>>101153936
Classic. A lot of the /aco/ folks use the Opera VPN method as well.
>>
File: 7886.jpg (1.12 MB, 1792x1024)
1.12 MB
1.12 MB JPG
>>
File: img_3.png (3.14 MB, 1792x1024)
3.14 MB
3.14 MB PNG
>>101154799
Somehow I see it as a metaphor of these threads status.

Good morning, anyway.
>>
>>101149608
How to generate widescreen images in DALL-E?
>>
>>101155115
In BIC/Designer, the old alternative was regenerating the pictures using BIC EDIT, it's linked in OP. But Designer now should offer to everyone a "Size" button where you can specify the resolution/AR.
>>
>>
>>
>>
>>
>>
File: (you).jpg (259 KB, 1024x1024)
259 KB
259 KB JPG
>>101152664
>>101138557

>This is /g/
No, it's include all 4chan's board, you samegen spammer. By the way, you spend a lot of time here.
>>
>>101146627
any tips for getting this nice painterly texture out of D3? when I ask for "paintings" the images I get have nice composition and colour, but they don't have a nice brush stroke texture like this. instead the zoomed-in texture looks noisy and nasty/grainy.
>>
>>101155633 (me)
>>101155619
this is an example of what I meant by the above, it's art but the texture is not painterly like the OP image, when you zoom in it's grainy and dotty.
>>
>>101155657
"painting with wide brush strokes" gets you a more paint-like texture.
>>
>>
>>
>>101155744
Even with "wide brushstrokes" the very strong effect visible in Mustachio's (tempera, I think) painting is a bit hit&miss on my end.
>>
>>
>>101155744
>>101155900
I see what you mean, your first image is what I want but the second one has the grainy texture again. But if it works at least some of the time that's good, thanks.
>>
>>101155968
If you're referring to my pictures, the one with the AK-74 was "wide brushstrokes, detailed tempera painting", while the following one with the gals dining was an earlier attempted I made before that suggestion was made by the other anon, "Heavy brushstrokes, detailed tempera painting by Yoshitaka Amano". I think "wide brushstrokes" does indeed result in something more similar to Mustachio's OP picture. Maybe later in the day he could confirm it.
>>
>>101155968
from what you're posting here, I'm guessing you think painting=thick layers of paint, and what you're after is called impasto.
>>
Is this one skirting too close to the sun or something? I asked for an adjustment and it didn't give me anything. I was running it via ChatGPT.
>>
File: 1719052820365493.jpg (120 KB, 1124x858)
120 KB
120 KB JPG
>>101146627
Don't they do this like every other week at this point? Has there been any actual breakthroughs in these lawsuits? Do they even win any of these?

nypost.com/2024/06/24/business/sony-universal-warner-sue-ai-startups-suno-udio-for-infringement/
>>
>>101156042
No, I'm talking specifically about the texture of the generated image when you zoom in and whether it's smooth and painterly vs noisy and grainy. I know what impasto is but asking it for it doesn't reliably eliminate the grain/screen door effect. It's an image model issue, not an art style issue.
>>
>>
>>
>>
>>101153942
I tried.
>A cartoon frog with red lips in Matt Furie's art style punching a fat bearded man wearing rimmed glasses and a fedora in the face
>>
>>101156369
Attempted with a Marvel shirt
>>
>>101156379
>>
Oh look, it's Moviebob
>>
>>
>>
>>
>>
>>101156229
>Do they even win any of these?
It's too early, I suppose.
>>101156369
My attempt:
>Heavy brushstrokes, detailed manga painting by Hirohiko Araki, JoJo style: meme frog comedically punching a fat young loser with a marvel shirt.
>>
>>101156461
Kek, that's a great one! I wanted to start out basic and see how far I could go. It even gave him a manbun!
>>
when doing multiple images at once (eg im doing left and right) is there a way to keep the prompts from bleeding into each other?
>>
>An angry meme frog punching a fat neckbeard loser with a stained Marvel shirt at his computer, heavy brushstrokes, detailed manga painting by Akira Toriyama, Dragon Ball style
>>
Well, it's not DB style, but it did come out comedically.
>>
>>
>>
>>101156461
A lucky one, too bad for the invisible hand passing through Pepe.
>>
>>101156654
Of course I meant to quote >>101156475

Back to the fluffy slop.
>>
#livingmybestlife
#6feetandupclub
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
File: 1719388378930186.png (543 KB, 910x660)
543 KB
543 KB PNG
Would someone here please post a prompt that would get Viconia DeVir from Baldur's Gate 3?
>>
>>
>>
File: 627.png (1.52 MB, 1024x1024)
1.52 MB
1.52 MB PNG
>>101151871
>>101153960
whoa nelly, it worked

it's not perfect, doesn't follow prompts quite as closely as bing (had to extend the bodysuit to cover her hand myself among a few other edits, though i'd always have been making certain ones anyway, like adding her tummy number), but the censor is WAY more lax. it's actually usable again!

in any case, it's so nice to finally see 627 again, thanks a bundle
>>
I really liked how this one came out. Also, this one was unprompted, but it looks like "crown braid" is working more reliably than how I remembered it. That's nice, I like that haircut (I blame youknowwho).
>>
>>101157148
Have you even tried it yourself? Nobody's going to spoonfeed you. Also, if you weren't aware, /v/ has a DALL-E thread too.
>>
This one was prompted. It doesn't work 100% of the times, but at least with my prompts even other kind of braids are hit&miss.
>>101158442
He tried that, to be fair.
>>>/v/680885642
>>
File: 7867.jpg (227 KB, 1024x1024)
227 KB
227 KB JPG
>>
File: 7863.jpg (214 KB, 1024x1024)
214 KB
214 KB JPG
>a painting in the style of "i'm out of ideas so put something really cool here kthx" by dorothea tanning with anime girls. watercolor style. anime style

It's out of ideas, too.
>>
>>
>>101159390
Fat cats and plump ladies? They were good for Boris Kustodiev and they are good for me.

>heavy brushstrokes, gouache&ink, detailed manga painting by Yoji Shinkawa, close-up, focus on face: Sicilian woman (cat ears and tail, fat, motherly, cheerful, loved-by-everyone, pudgy chest, long straight black hair, fringe, payots, blue eyes, fair skin complexion, colorful Sicilian traditional attire) holding on her laps her fat fluffy black cat. Sitting on a chair near a window of her house in Palermo. Spring, clear sunny day, night, full moon.
>>
>>101160062
Yeah, I forgot to remove the sunny day bit from the prompt that I modified... make it clear sky or whatever.
>>
>>101156489
Probably just luck
>>
anybody know how to replicate the general tapering shape of this ass? picrel.
prompt I'm using:
>dynamic zoomed out shot of a woman drinking coffee, on a city street leaning on the hood of a police cruiser, wearing a simple traditional shrine miko robes turned into tight police uniform, messy hair, epaulettes on shoulders, painterly, scratchy, rough strokes, flat illustration style, vivid highlights and shadows
>>
>>101156212
dunno, what's the prompt? i'll try it out
>>
>>101151518
kek
>>
>>
>>
>>101161328
>young admirer protects necromancer from her own conjurations
>>
File: 7896.jpg (1.71 MB, 1792x1024)
1.71 MB
1.71 MB JPG
>>
File: bing 1719425103663.jpg (62 KB, 1024x1024)
62 KB
62 KB JPG
Hello
>>
File: bing 1719425773549.jpg (81 KB, 1024x1024)
81 KB
81 KB JPG
>>
>>101163228
Peak AIslop.
>>
File: 00040-65505437.png (1.81 MB, 1024x1024)
1.81 MB
1.81 MB PNG
>>
File: 1696272439628615-1.png (1.02 MB, 1024x1024)
1.02 MB
1.02 MB PNG
>>101163716
>>
File: final drow.png (1.64 MB, 2834x3967)
1.64 MB
1.64 MB PNG
>>101163750
>>
File: OIG1.jpg (318 KB, 1792x1024)
318 KB
318 KB JPG
16mm photography with surreal cross processing printing. An house near a beach, sunny day.
>>
File: OIG2.jpg (170 KB, 1024x1024)
170 KB
170 KB JPG
How do I stop the AI from using anime style? There's no negative prompt in this
>>
Wanted to animate some stuff in Luma today but it appears it's gotten even worse. Oh well
>>
File: OIG2.jpg (178 KB, 1024x1024)
178 KB
178 KB JPG
>>101163806
>>
>>101163883
>>101164455
>when she tires of your incessant faggotry and readies her favorite ice spell to attack

try "western illustration"
>>
File: 8cfug.jpg (482 KB, 1024x1024)
482 KB
482 KB JPG
>>
>>101164455
this beauty looks so calm yet fierce/strong, absolutely delicious
>>
>>101163716
>>101163750
>>101163806
>>101164455

Good shit anon.
>>
File: llllll.jpg (233 KB, 1024x1024)
233 KB
233 KB JPG
>>101165467
>>
File: bing240626-154307_7641.jpg (221 KB, 1024x1024)
221 KB
221 KB JPG
>>
File: fbae7.jpg (169 KB, 1024x1024)
169 KB
169 KB JPG
>>
File: 7910.jpg (554 KB, 1792x1024)
554 KB
554 KB JPG
>>
>>101156212
this seems like it has some potential, care to share what prompt you used?
>>
>>101150806
god i wish that were me
>>
>>
>>
anybody know how to prompt this kind of swept hairstyle consistently? this showed up randomly with a messy hair prompt, I really like it
calling it sideswept or assymetrical doesn't replicate it
>>
File: OIG4.2RVOx.jpg (119 KB, 1024x1024)
119 KB
119 KB JPG
You're here to buy some narcotics right? Could you say it very clearly into my coat?
>>
>>101167968
Windswept?
>>
>>101167968
Messy rooster?
>>
>>101168268
>>101168411
no dice on either
>>
>>101167431
Hahaha.
>>
File: OIG1.jpg (198 KB, 1024x1024)
198 KB
198 KB JPG
>>101165617
>>
File: OIGff3.jpg (136 KB, 1024x1024)
136 KB
136 KB JPG
>>101168659
>>
how do you get busty girls? dall-e freaks out if i mention anything about a woman's stature (or really anything about a woman)
>>
File: OIG4.DKEj.jpg (153 KB, 1024x1024)
153 KB
153 KB JPG
>>101169322
Luck. If you really just want to make porn DL and learn stable diffusion
>>
File: Designer(3).png (1.92 MB, 1024x1024)
1.92 MB
1.92 MB PNG
>>101169322
foxgirls are totally okay, until they want me to stop
>>
>>101168659
>She tries to pick up humans at a bar
>She keeps trying to work the word "horny" into her lines
>She nervously jokes "Get it! Because horns!", while pointing at her head.
>Frustratingly her efforts are going unrewarded.
>Retreats to her booze, "Pearls before swine. I gave you comedy gold!"
>>
>>
>>
File: Bing .jpg (145 KB, 1024x1024)
145 KB
145 KB JPG
It seems AI eventually will always default to ignoring the user's input, then now Dalle also ignores the user's input for something it thinks you want instead, exactly like youtube and google search.
>>
File: robo.jpg (201 KB, 1024x1024)
201 KB
201 KB JPG
>>
>>101171597
Youtube is actually very good at suggesting me stuff that makes me become depressed for several days. *Open an innocent thumbnail of a bobcat without reading well the title. It's a fucking trapper teasing the beast before killing it. FML.*
>>
>>101155144
>BIC EDIT
it's not working for me, at least the 1.7 version
>>
File: OIG2.jpg (280 KB, 1024x1024)
280 KB
280 KB JPG
>>
File: futuristic lesbians.jpg (128 KB, 1024x1024)
128 KB
128 KB JPG
>>
File: cake decoration.jpg (215 KB, 1024x1024)
215 KB
215 KB JPG
>>
File: OIG1.hoHwXo2R8LzK4KYDd.jpg (200 KB, 1792x1024)
200 KB
200 KB JPG
>>
>>
>>101172511
Try this:
https://desuarchive.org/g/thread/100914325/#100928545
>>
>>101172864
no luck
>>
>>101173333
Another anon on /v/ similarly had no luck yesterday, I guess it might be a new issue; my (obvious) suggestion is using a different account for now.
>>
File: 7934.jpg (318 KB, 1024x1024)
318 KB
318 KB JPG
>>
>>
File: 2024-06-27.jpg (206 KB, 1024x1024)
206 KB
206 KB JPG
>>
File: 1718992467160.png (2.9 MB, 1792x1024)
2.9 MB
2.9 MB PNG
>>
Sexy handsome bump.
>>
>no sex
>no violence

what does microsoft think people like
honestly
>>
>>101177271
Animals? Lol.
>>
>>101177296
ironically I am attempting to make some horses now and getting blocked
>>
File: OIG1.x0sO48tdFVnxlCF_y.jpg (213 KB, 1792x1024)
213 KB
213 KB JPG
>>
>>
>>
I was trying to visualize my anger
>>
>>101177378
>I was trying to visualize my anger
Try harder.
>>
>>101177449
Nah I'm alright now
>>
>>
>>101177592
>>
>>101177271
honestly, i would learn to use a local model, but i've heard that SD3 is garbage
>>
>>101157038
Care to share the prompt ?
>>
>>
any way to gen nice girl abs consistently
their bar for filtering is so low that any sort of "fit girl" and "abs" gets dogged
>>
>>101173379
4 accs no dice
>>
>>
File: e4c.jpg (549 KB, 1024x1024)
549 KB
549 KB JPG
>>101177378
gigga nigga
>>
>>101178647
Athletic? Sporty?
>>
>>
>>101178647
Mention "6pack" somewhere to help get abs.

The trick is, men have abs and generally show their nipples, so the gen is putting in nipples and getting dogged. You will need to ensure that her top is drawn.

By default, the AI sees her top as a solid field of color from her neck to her waist. You have to tell it otherwise. What you really need, is a visual horizontal line to separate her chest from her midriff. It can help to specify that the top is "short" or "tattered" or "tied".
>>
>>
New.
>>101182014
>>101182014
>>101182014



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.