[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
/gd/ - Graphic Design

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • Additional supported file types are: PDF
  • There are 94 posters in this thread.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


Janitor applications are now closed. Thank you to everyone who applied!




Last thread hit the image limit, Taking requests for anything anyone wants to have created. Nothing NSFW or related to politics/public figures (site doesn't allow). As a general rule of thumb if you can't imagine it in your head, dall-e 2 can't either.
>>
This image was made from "Cowboy sitting on horse, fishing in river, snow peaked mountains, sunset, clouds, western style painting" so maybe that gives you guys a starting point of what to write.
>>
Black hole demon with purple eyes
>>
Cartoon style Nikocado Avocado eating a can of sardines
>>
Satania from Gabriel DropOut giving a computer science lecture to a bunch of neckbeards
>>
“Western style painting of cowboy sitting on log putting hands towards campfire with hunting dog laying beside him, night time, stars, moon”
>>
“Cowboys and cowgirls waltzing in barn with country music band on stage, sun light beams shingling through, 1800’s, western style painting”
>>
baroque painting of a war scene in outer space
>>
The golden knight, bloodstained, sitting on top of the dragon he slain.
>>
>>433039
Dark Souls gothic mega city painting.
>>
File: wfuijavfw.png (546 KB, 1039x693)
546 KB
546 KB PNG
>>433047
>>
File: yawvfohuawfgll'.png (744 KB, 1043x694)
744 KB
744 KB PNG
>>433056
ok this ones kinda funny
>>
File: 89wadyhuioag.png (770 KB, 1042x696)
770 KB
770 KB PNG
>>433055
Since this is early in the thread, a reminder to everyone to keep it simple. It doesn't know a lot of internet culture including video games. Also pls keep your prompts interesting
>>
File: 7ta8wguyfihbvj.png (1.17 MB, 1044x695)
1.17 MB
1.17 MB PNG
>>433057
bottom middle is nice
>>
File: 7tag9uyhibjaw.png (1.37 MB, 1037x692)
1.37 MB
1.37 MB PNG
>>433062
>The golden knight, bloodstained, sitting on top of the dragon he slain.
>>
File: 89tpguiojbag3.png (1.25 MB, 1036x704)
1.25 MB
1.25 MB PNG
>>433063
good prompt
>>
Cat with a wizard hat, renaissance style painting
>>
File: 8tguboj laf.png (1.24 MB, 1045x702)
1.24 MB
1.24 MB PNG
>>433073
>>
Brutalist noir skyline at sunrise in the style of Wiktor Sadowski
>>
Pizza topped with sardines
>>
>>433040
Next time leave out "painted by somebody's aunt who gives paintings as gifts to people who immediately donate them to a thrift store where they languish for years with all the other terrible attempts at creating art that nobody else want either" and see what you get.
>>
>>433039
Devastated futuristic city with ruined sky scrapers all around, smoke rising from the ruin, brown sky
>>
Eye: Divine Cybermancy
>>
Human Silhouette falling on a giant multicolor vortex painting
>>
>>433039
Sky falling like a broken mirror over a mega city VHS video
>>
Jaguar surrounded by a millon blue morpho butterflys on a photo studio
>>
desert at night door portal to out space stars
>>
>>433069
>>433039
>As a general rule of thumb if you can't imagine it in your head, dall-e 2 can't either.
>keep it simple
All premised on a notion that there is some ideal response to these prompts that can be attained by tailoring them to what you think that ideal might look like.
Nothing wrong with that approach where it's appropriate but it's very heavily weighted towards an engineering/ problem solving approach that avoids the random surprise and unstructured expectation aspects of art that differentiate it.
IOW while both art and engineering/ design are derivative to some extent and require selection of goals and experimenting and vetting of results to see if they fit, the fact that AI can't really think or reason actually makes it well suited for an approach that treats overthinking as something to avoid.
That is to say that not helping it with streamlined instructions and deliberately confusing it with vague prompts and nonsensical or conflicting terms can yield great results too and the nature of computing means it isn't hampered by ego or fear of coming up with the "wrong" response.

Pic related: "Chuck Berry Breakfast Donkey" via Wombo Dream.
>>
>>433093
Also fun to use non-visual and conflicting prompts to see just how inside-the-box AI responses are, like "silent hot loud antique" which Wombo obviously "creates" by combining some pretty on-the-nose tropes like big empty room for silent, fireplace for hot, a stereo for loud, and everything brown and/ or dusty for antique.
It's still kind of interesting but not that different from what a lazy and not too creative artist might render.
>>
“Cowboys and cowgirls waltzing in barn with country music band on stage, sun light beams shingling through, 1800’s, western style painting”
>>
>>433061
"Who was Frank Frazetta?"
>>
>>433039
Sentient insect in giants hollow heel

Every butterfly to ever exist
>>
>>433039
"Cowboy sitting on horse, by river, American West, oil painting by William Ranney." (Tylee is middle name if that doesn't work)
>>
File: 6r8ofycghv jbn2g4.png (285 KB, 1042x694)
285 KB
285 KB PNG
>>433076
>>
File: 6ryfuvhgjag.png (1.06 MB, 1044x706)
1.06 MB
1.06 MB PNG
>>433086
good prompt
>>
File: 6rfyvhag.png (1.43 MB, 1056x699)
1.43 MB
1.43 MB PNG
>>433088
>>
File: 5yfugchaf.png (1.61 MB, 1055x701)
1.61 MB
1.61 MB PNG
>>433078
very realistic
>>
File: 76a8y3fuhvafy8uvahf.png (1.56 MB, 1051x694)
1.56 MB
1.56 MB PNG
>>433089
>>
File: 6fyuvhag.png (1.03 MB, 1045x693)
1.03 MB
1.03 MB PNG
>>433093
Nah I get what you're saying but there is something wrong with it bc I get bored of people trying to do "the meaning of life" and then getting this type of shit lol
>>
File: w45srdtxcfgh .png (1.1 MB, 1051x704)
1.1 MB
1.1 MB PNG
>>433092
>>
File: 5ed6tycghjag.png (1.47 MB, 1046x689)
1.47 MB
1.47 MB PNG
>>433096
>>
File: e546drtycfghafw.png (1.42 MB, 1037x691)
1.42 MB
1.42 MB PNG
>>433098
>Sentient insect in giants hollow heel
>>
File: 5erf7tvuh.png (1.41 MB, 1041x697)
1.41 MB
1.41 MB PNG
>>433100
>>
>>433039
White and gold exotic buildings on a large cliff overlooking the ocean, waterfalls coming from the cliffs, another planet can be faintly seen in the blue sky
>>
File: 678yfiuhvjat3y.png (1.36 MB, 1042x700)
1.36 MB
1.36 MB PNG
>>433116
>>
>>433106
Yeah, I get that but a lot of the prompts people do come up with are so specific and all encompassing that there's no room for any novelty or surprises. They remind me of clients who come in with a laundry list of elements they want in some design that essentially leave no room for creativity and design.
They might as well just be clip art or photoshop; " I want a goat, and a car and a phone booth and a blimp at sunset with mountains" isn't hard at all to imagine or compose and the incongruity doesn't really serve any purpose when so little is left to interpret. You could literally use props to create that scene.
There's a fine line in between where you could get some interesting results but the ones in this thread are pretty mundane and unimpressive.
That was my only point, that stuff like this requires rethinking your entire approach to get anything that different than what that approach already allows, and the approaches I see are the source of boring results.
This also applies to the approach developers take to designing the processes, which is often nothing like how exciting art and design are created.
>>
Evil personification of the covid vaccine
>>
Glowing pink colored heroin on a table in a dimly lit room with a crucifix hanging on the wall
>>
Gorilla playing video games
>>
File: 78fiyugchva.png (1000 KB, 1038x697)
1000 KB
1000 KB PNG
>>433122
covid isn't allowed
>>433124
no drugs
>>433128
here is a photo of a gorilla playing video games
>>
Saturn Devouring His Son calarts style
>>
Red eye cyclops with flaming sword sitting in his throne
>>
Shenron from DragonBall z but white
>>
File: e5dtcyg khljn;kl.png (673 KB, 1048x717)
673 KB
673 KB PNG
>>433135
>>
File: esxrfcyiguvhbijk.png (1.32 MB, 1045x695)
1.32 MB
1.32 MB PNG
>>433136
>>
File: 4sr6dctiguvh'.png (772 KB, 1081x709)
772 KB
772 KB PNG
>>433137
>>
Europa the lazy in her awakend form from claymore at a tea party with Alice from alice in the wonderland and Batman
>>
Could you please do:
Graphic Design Logo for Car Washing Service called ISA using the colors blue and yellow
>>
>>433138
LOFL, utterly terrifying capacity for intuition and problem solving.
Worrying about this shit is like worrying if Mad Libs is going to put writers out of business.
>>
>>433149
>worrying if Mad Libs is going to put writers out of business.

no, A.I. will do that as well
>>
File: 687rfyuvhgaf.png (1.51 MB, 1041x703)
1.51 MB
1.51 MB PNG
>>433146
>>
File: 6r7ftcgawgf.png (384 KB, 1050x701)
384 KB
384 KB PNG
>>433147
>>
>>433152
Kek, see >>433157 , with only three letters specified it went 1 for 6 .

Not gonna hold my breath.
>>
Hyper realistic dragon swallowing a blue sun
>>
Scifi battle tank with Grey armor and an energy weapon
>>
File: tguivhj.png (1.93 MB, 2349x601)
1.93 MB
1.93 MB PNG
>>433165
against tos for some reason
>>433164
>>
Sci-fi marine's helmet dark gray and red with a purple visor
>>
Cyberpunk dwarf tinkering with a robot's parks in a rundown workshop
>>
Great Iron airship sails into the sunset in the style of starry night
>>
Classical painting of knights riding through the woods
>>
Red and white logo of an asteroid mining company
>>
>>433157
Kek thank you very much anon appreciate it!
>>
50mm commercial shot of an ancient set of black plate armor with purple rune inscription along the edges
>>
>>433039
physical representation of entropy in a 3d form
>>
Smiling Angolan Chimpanzee Boy
>>
>>433039
why is the Dall-e 2 team pushing it so hard on this dead board
>>
File: 57tqgyfvh .png (928 KB, 2368x627)
928 KB
928 KB PNG
>>433180
>>
File: 76t8ygihblfaw.png (2.74 MB, 2358x601)
2.74 MB
2.74 MB PNG
>>433181
don't do that many prompts pls
>>
File: 6rf8oyvuhlajwg.png (2.39 MB, 2358x603)
2.39 MB
2.39 MB PNG
>>433188
here
>>433198
against tos
>>
File: 6fyuvhj.png (1.73 MB, 2334x643)
1.73 MB
1.73 MB PNG
>>433195
>>
File: 56r7tfugvafw.png (2.4 MB, 2360x607)
2.4 MB
2.4 MB PNG
>>433182
>>
>>433183
>>
The man who invented peanut butter
>>
>>433179
Kek, renders four blue dragons and only one blue sun with the other three yellow, none of the elements look realistic at all let alone hyper realistic.
>>
>>433039
I know this is long, but here goes:
woman in futuristic armored swimsuit with glass bubble helmet and boots and long gloves holding raygun fighting brains in space 1930s 1940s 1950s pulp magazine science fiction Earle Bergey
>>
>>433219
there are no images of hyper realistic dragons lol
>>
>>433222
>there are no images of hyper realistic dragons lol

You do realize that something doesn't have to have actually existed in real life to be rendered in a hyper realistic graphic style, don't you?
>>
File: 5rt6cygivhjafw.png (34 KB, 577x369)
34 KB
34 KB PNG
Going to call it here since now I got to pay lol
>>
>>433039
The black hole general fights the one eyed octopus sun, riding a platinum pegasus over vapor and rainbows
>>
>>433039
inkblot test
>>
>>433234
>13 cents per image gen
I'd do it for 3-5c, but I'll still take my free 15 and maybe they can hook me with that.
I assume costs will come down eventually, especially once proper competition opens up and they fully scale.
>>
>>433159
As someone who has played with these models for a few years, If I got those results 18 months ago my head would have exploded.
>>
>>433286
Meaning what? That 1 for 6 following a specific prompt is better than before, or worse?
>>
>>433105
can i ask for 4th?
>>
File: hummingbird-guitar.png (299 KB, 1709x660)
299 KB
299 KB PNG
>>433287
Yup, adherence to prompts has improved radically. Letters showing up at all used to be out of the question, so I cannot complain when letters get jumbled up.

Additionally, if you look at other SOTA models like Parti, or Imagen, shorter texts are actually capabilities right now.
>>
>>433234
If anyone here wants 30 free DALL-E 2 credits, my friend hosts a prompt engineering contest where that’s the 1st place prize. It’s a 1v1 bracketed tournament, where you’ve got 2 minutes to send in your prompt every round, and watch it generate live versus an opponent’s prompt. Funnest and most interactive way I’ve found to interact with DALL-E 2, as someone without access.
>>
Prompt:

Realistic image of Heavy from team fortress 2
>>
>>433075
I loved all of them

>>433103
For some reason it remind me early 2000's futuristic art. Like Deus Ex invisible war, or like the atii short with circuits and new york style buildings ,i forgot its name

>>433107
Loved the ones on right, i wish i could download them one by one

>>433117
Loved all of them except those on the top right and top left

>>433212
Disappointment

>>433387
Downloaded it
>>
>>433387
That tournement it as an awesome idea, looking forward to tuning into the next one
>>
Man, now that Dall-e 2 is charging for credits, I wish you could just buy it instead of being put on a wait list.
I have money, and I'm willing to spend it on this.
>>
Anyone do something like:
Coco-Cola bottle shaped like a polar bear set on snow.
Or
Decanter filled with prismatic fluid shaped like a unicorn. Realistic.

I might be a bottle collector...
>>
Castlevania SOTN, Cinematic, Concept art
>>
>>433380

Thanks for the clarification

>Additionally, if you look at other SOTA models like Parti, or Imagen, shorter texts are actually capabilities right now.

Again, not quite clear what you mean..."shorter texts" meaning it can accurately create what short combination/order of letters is asked of it , like IBM or USA or MSNBC?

Progress is progress, I guess but compared to what humans can do in this respect I still don't see where it is anything to fear as far as taking work from people, or will be at some point in the future.
As an approach to design/ composition it doesn't seem to offer any new or novel perspective and just mimics the way artists already do this, with large gaps in its ability to stretch beyond that built in boundary of "this is the process by which you achieve X", with X being some ideal design/composition.

Sure, lots of graphic design works that way but it can already be done with basic templates and formulas that don't require much intelligence at all to " create" with.
The more creative any designer is, the more he steps away from rigid protocols and formulas as the final arbiter of what to turn in, while holding to protocols that need to be adhered to like putting the correct letters in order in a logo design with no extras.
What I see is the exact opposite.
>>
File: pardi_model_size_comps.png (1.4 MB, 1645x782)
1.4 MB
1.4 MB PNG
>>433481
>Again, not quite clear what you mean..."shorter texts" meaning it can accurately create what short combination/order of letters is asked of it , like IBM or USA or MSNBC?

Yep, as the models have gotten larger, the the accuracy of letters showing up has increased as well.

>doesn't seem to offer any new or novel perspective
>The more creative any designer is, the more he steps away from rigid protocols and formulas as the final arbiter...
I think you put these paragraphs very well. I won't disagree presently, but hypothetically what would you need to see from a text-to-image model to be convinced otherwise?
>>
File: bottles.png (1.91 MB, 1359x964)
1.91 MB
1.91 MB PNG
>>433468
Pick any you like and I'll blow em up

>>433472
https://labs.openai.com/s/egMk1rQmhSjsdHojK2kYDsQP
https://labs.openai.com/s/szvZlif2do27uQAm7DAW8TB2
https://labs.openai.com/s/EEPjiZFDDEV3yw3kY1eS6eUu
>>
>>433071
Surprised that bloodstained isn’t a blacklisted term
>>
>>433498
I'd say of the first four, 2 and 3 are the best, and of the bears, 1 and 3 are the best.
>>
>>433501
Though of the bears, I should say that 7 was the closest to what I was thinking of.
Did you know that Coca Cola never made a glass bottle shaped like a bear despite them being the mascot for so long?
Like, I've always imagined a bottle similar to honey bottles, you know the ones, plastic, shaped like a bear with the nozzle coming out of the head.
Seems like a no-brainer to me.

Also, image 4 of the bears, while the furthest away from the prompt, is probably the best picture, still.
I have no idea what happened to number 8, though. Looks like Dall-E had a stroke.
>>
>>433496
>hypothetically what would you need to see from a text-to-image model to be convinced otherwise?

Honestly, some proof that there was some need for it or reason for trying to do it in the first place.
I do see some value in a tool that gets you out of old habits and scripts and have played around with some stuff online but more in the sense of trying to shake it loose from what it's "supposed" to do. That to me is the hurdle to any creative endeavor; scripts and formulas and recipes have their place but tweaking and re-imagining them is where the creative spirit manifests.
When it comes specifically to text, labels and the like that becomes less valuable and familiarity ie being able to interpret the text is crucial.
It doesn't really matter to me that these AI systems get mixed up on text, the point is why does anyone need AI to spell or lay out text in a graphic design based on someone's prompt to begin with?
It's dirt simple to begin with and what is the practical impetus for creating artificial " intelligence" to do it?
It's amazing and all but other than as a means of researching AI or selling the next great thing to people who care about that, I really don't see what advantage it has that would displace human designers any more than all the dumb methods like clip art and design templates and stencils and photography and their digital counterparts did before AI.
It's like self driving cars- there's a place for them but much of the hype around them is a solution looking for a problem.
People aren't clamoring for them because doing it the old fashioned way isn't that hard and they don't want to hand over total control to a machine.
>>
>>433496
>I think you put these paragraphs very well.
Thanks, it's something I've given a lot of thought to when creating art/ design but also music that tries to avoid becoming stale and repetitious and predictable while remaining relatable. Musicians have been using computers and electronics to assist them for a long time, long enough for many cycles of novel approaches becoming trite and clichéd.
One cool tool that may seem off topic but imo really isnt is the Oblique Strategies card system that assists in breaking through blocks and rote approaches in human creators -

"Each card contains a gnomic suggestion, aphorism or remark which can be used to break a deadlock or dilemma situation. A few are specific to music composition; others are more general. For example:

Use an old idea.
State the problem in words as clearly as possible.
Only one element of each kind.
What would your closest friend do?
What to increase? What to reduce?
Are there sections? Consider transitions.
Try faking it!
Honour thy error as a hidden intention.
Ask your body.
Work at a different speed..."
https://en.m.wikipedia.org/wiki/Oblique_Strategies
Note that the second suggestion
>State the problem in words as clearly as possible
is essentially the basis for AI art programs...its a valid approach but it's only one approach and in that system it's no better or worse a suggestion than "ask your body".
Ultimately the human decides what works and what doesn't and the ability to make those determinations and benefit from even silly or inscrutable suggestions like "Think of the radio" is where AI can't compete.
1/2
>>
>>433509
2/2


One of the Oblique Strategies prompts I like a lot is "Ask people to work against their better judgement" and that kind of sums up how I approach all of this stuff, I want to see and hear raw material and concepts that aren't filtered and constrained and tempered by formula or reason or even intelligence, and then refine them as needed into completed works.

Anyway, here's an online version, imagine being able to get this stuff in an AI artist's head along with a prompt-

http://stoney.sb.org/eno/oblique.html
>>
A classical drypoint etching of bigfoot and a medieval knight fighting on a large field in the style of Giuseppe Arcimboldo
>>
>>433544
https://labs.openai.com/s/ItI4xzN2eU0QoVBfyO5hT3j1
https://labs.openai.com/s/Ahwmg7c5pbsksremO5mMfC63
https://labs.openai.com/s/k6GDUuM3bweIq1K5mrhM8TpM
https://labs.openai.com/s/A3IOBSL8TVX3fHyvvQKc4Qb8
>>
File: parsnip 001.jpg (358 KB, 630x1200)
358 KB
358 KB JPG
>>433547
Wow, lower left hand corner looks just like a photograph of bigfoot, eerily accurate
>>
>>433547
Here's the same prompt with craiyon
>>
>>433039
So...
You're getting people to type in what they want from an AI image generator, take that to the image generator, enter it there, get the image, and put it in a reply to the original poster??
Are you a government efficiency advisor?

And why don't the rest of you just get off your asses and do it yourself?
Here's a link to the most recent release (now crAIyon)
https://www.craiyon.com/
>>
>>433568
Here's an example.
It's pretty easy.
>>
>>433568
>>433569
imagine being this smug having no idea how dumb you really are
blessed life t bh
>>
>>433569
These are made with Dall-e 2, which is invite only. And so far the alleged most advanced AI generator for images.
>>
>>433568
craiyon sucks
>>
>>433223
i made a realistic dragon head for you lmaooo god its so ugly haha
>>
one I did yesterday. wont take any requests but I can share more later tonight if anyone is interested in seeing them.
>>
>>433569
gotta love how "people finding the least efficient way possible to complete a simple task" returns an entire row of those generic "teamwork team working together as a team!" stock pics corporations love to slather all over their PR materials, where "diversity" is always consideration #+1 and always means statistically overrepresenting women and minorities.
Wry joke/ commentary as the AI develops, or just the GIGO principle in effect?
I'd say the latter based on how many times I've had search engines return those "diversity team" images for searches that had only the weakest and most tangential connection to that kind of thing.
>>
>>433591
Yves Tumor
>>
>>433642
>>
>>
>>
>>
>>
>>
>>
>>
last one goodnight all
>>
File: severian aureole.png (1.22 MB, 1024x1024)
1.22 MB
1.22 MB PNG
How does Dall-e2 compare to Midjourney?
>>
An adventurer in a fantasy world town wearing Monarchy on his head.
>>
"Pepe the frog drives a moped through Pompeii"
>>
>>433039
Do you have unlimited credits or something? How/why you doing this if it costs the moolah?
>>
>>434332
>Do you have unlimited credits or something?
>How/why you doing this if it costs the moolah?
If you get accepted in their beta (and it's easier to be visited by Jessica Christ in a wet dream) you get 50 free credits and after that they'll give you 15 free credits per month. 4 image generations per credit, 3 image edits per credit, and 1 image enlargement per credit. If you were lucky enough to get in the beta early they'll reward your luckiness with an extra 100 credits.
But there's a thing, those 65 credits you get at the start vanish after a month whether you use them or not, so people could just fulfill requests when their credits are about to disappear, anyway.
>>
Batman fight scene painted by Salvador Dali
>>
My favorite thing is making old Rembrandts.
>>
>>
>>
>>
>>434358
Not OP, but I'm bored
>>
>>434378
And another
>>
File: 1657852817702.png (21 KB, 864x836)
21 KB
21 KB PNG
Gosh I miss the kind of requests that were being done a month ago in the /vg/ section of 4chan, people would send picrel as request to be inpainted...

(cont)
>>
File: 1657857675716.png (1.52 MB, 1024x1024)
1.52 MB
1.52 MB PNG
(cont)

Then someone would use dalle 2 to inpaint it and post the result.

(cont)
>>
File: 1657858045076.png (1.48 MB, 1024x1024)
1.48 MB
1.48 MB PNG
(cont)

And then someone would post a dalle variation like picrel, and all of this was awesome. I thought the golden days of dalle request were over when credits came, but no, people still fulfill requests, the days are over because the requests have gotten lame...
>>
>>433156
bottom middle is actually beautiful
>>
amateur photo of a goth couple wandering Yosemite on a cold, cloudy day
>>
File: goth_couple_yosemite.png (2.83 MB, 1408x1024)
2.83 MB
2.83 MB PNG
>>434421
Attached pic is StableDiffusion

Links are Dall-E 2
https://labs.openai.com/s/f8rgkFEAvmBWaxlIn1s627uy
https://labs.openai.com/s/qhQVwMKtNhoOruijiC1q3MYt
https://labs.openai.com/s/LOucNCN1a1CYGOHKFPAVpr33
https://labs.openai.com/s/rqXjc4Xol4nNQllYJrUnULgF
>>
File: emma.gif (4.87 MB, 1200x1200)
4.87 MB
4.87 MB GIF
The seed specification parameter in StableDiffusion makes editing prompts incredibly addictive. In terms of visual effect when cycling through, and as a helpful tool to see what keywords make what effect.

It seems like it would be a no brainer for the folks at OpenAI to include with Dall-e 2.


This gif of SD outputs are Emma Watson in: RDR; The Last of Us; NBA 2K; Call of Duty; Fortnite; Grand Theft Auto
>>
>>434235
Midjourney is cool but has nothing on dall e 2
>>
>>434235
I've been enjoying Midjourney, imo it seems to give more creative control than Dall-E. Dall-E's images are typically more coherent, but I got bored of it fast.
>>
>>433039

a post punk album in style of kazimierz malewicz
>>
>>434466
album cover, damn
>>
>>434466
Dall-E 1/2
>>
>>434471
Dall-E 2/2
>>
>>434466
Midjourney 1/2
>>
>>434473
Midjourney 2/2
>>
>>434474
>>434473
>>434472

Jesus that's excellent, the second Dall-E one even has registration marks at the bottom, simillar to that New Order album
>>
>>434454
>it seems to give more creative control than Dall-E
it's the opposite imo. midjourney has progressively gotten more of a distinct look, where, at the default settings of v3 the user has very limited control over the actual style of the image.
the images look better yes, but because the model is more rigid. you have less creative control and the results are better precisely because of it.
>>
>>434525
midjourney most samey. dalle-2 has its own "look" too. SD has the most variable and malleable results, you can get some WILD shit. it's like craiyon on steroids
>>
>>434526
a lot of dall-e 2's limitations feel like intentional gimps to prevent controvesy. on similar prompts where they both produce expected results, the images from dall-e are definitely a lot more coherent than SD, higher fidelity etc.
guessing most of that is just a result of the model being 12 gajillion parameters large.
>>
>>434525
Midjourney only looks same-y because everyone uses the same 3 descriptors for each prompt. The AI can only be as creative as it's user's allow.
>>
Ancient sith warrior on dusty red planet
>>
>>434526

Are you using SD through the diffusers library or just the github original? Also, is there any point in downloading any model checkpoint aside from v-1-4 or is it near enough "strictly" better than the others that you'd only ever vary the prompt/options and never try an earlier checkpoint? I notice it's an extension of 2, while 3 is just the same as 4 with fewer steps and no training on aesthetics-v2
>>
Midjourney's officially mogging every other AI
>>
>>433039
Skateboarder using a skateboard to hit a surprised member of the ku klux klan lying on the floor
>>
Warwick Davis floating among atoms and molecules
>>
File: AncientSith.jpg (1.97 MB, 2048x2048)
1.97 MB
1.97 MB JPG
>>434556
Here you go my bro.

>>434526
SD runs on my computer and gives pictures in the ballpark of Midjourney and DALLE2 in 10 seconds (needs good descriptors tho) and it's fucking insane, I can't believe it, we can finally get rid of arthoes
>>
File: midjourney.png (452 KB, 512x512)
452 KB
452 KB PNG
>>433040
Same input on midjourney
>>
Evil Cat Eating Red Bellpepper

maybe in a cartoon / painting style. Maybe the Evil Cat wearing Devil Outfit ?
>>
>>433068
>ashg trake
>Sporeher chiat SFT
>tarıta Scotcchetar
>Gata shatieia

What did she mean by this
5
>>
Still taking requests?
>>
>>433039
A face of gold on a blue gradient background
>>
Anyone know where to find a guide for installing dalle locally?
>>
>>435045
use stable diffusion
rentry.org/guitard
>>
File: dalle2.png (2.22 MB, 1188x1292)
2.22 MB
2.22 MB PNG
guys...
>>
>>435099
can you even imagine how terrifying the first "ai movie" will be
>>
>>435101
What if future AI could make its own feature-length movies based solely on a given script, which itself was AI-generated?
>>
>>434815
>>
>>435039
>>
obese decaying corpse of a man being disassembled surgically by horrifying humanoid creatures
>>
Future person surfing through happiness meeting God
>>
>>435374
>>
>>435294
Idfk
>>
A realistic oil painting, a man standing in the middle of a solar system, light sparkles, crystal colours illuminate from within, light splashes and mixes, dragons circle around the rim of the painting, spaceships fire Lazer beams and destroy cities.
>>
flanged iccreamery staffed by ccthulhus
>>
Vampire kid stomping on his slain enemies, Murakami psycheldic horror
>>
Posting the best of my DALLE-2 results
>>
>>435461
>>
>>435462
>>
>>435463
>>
File: Science n Hobbes.jpg (37 KB, 600x273)
37 KB
37 KB JPG
A colossal sandworm rising before the sun.
>>
Do “Future person surfing through happiness meeting God”
>>
The eye thinking.
>>
A wrinkle in the out.
>>
Not gonna repost so soon, but I've been fucking with running Stable Diffusion on my own artwork and having some fun results. From the other thread, preamble + result.
>>435614
>>435629
>>
>>435156
Put the script of Sunspring through it.
>>
Hugh Neutron In the style of Leonardo Da Vinci's the Vitruvian Man
>>
Ancient text in middle east containing secrets of small hat tribe domination of humanity
>>
The more prompts I see written out the more convinced I am that the point at which AI will be able to do anything one spells out for it to do will coincide perfectly with the point at which a societal state of universal and absolute illiteracy is achieved.
>>
File: th-2609905726.jpg (24 KB, 474x266)
24 KB
24 KB JPG
>>435760
We're there already, mate. I saw a billboard advertising an "anti-dehydration gel", basically a chemical cocktail not dissimilar to pic related. If that is how stupid the average person is, total illiteracy and A.I takeover are just around the corner.
>>
File: file.png (631 KB, 512x512)
631 KB
631 KB PNG
>>
File: file.png (619 KB, 512x512)
619 KB
619 KB PNG
>>
Cartoon of can sitting in lawnchair with sunglasses on watching a nuclear blast
>>
Bouba and Kiki

https://en.wikipedia.org/wiki/Bouba/kiki_effect
I doubt it'll produce something interesting but just curious.
>>
>>
Jetsons concept art
>>
>>433039
Xi Jinping in front of Australia and Canada flag in ((painter)) art style.
>>
>>433039
Beautiful gothic elf queen in dress

Make it portrait style and very realistic. Wallpaper sized.



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.