images-discussions

[shut niche]

I want to adopt a snow monster and a wolf now, lol.

[late blade]

oh ya, we have a similar talk with the crimes @glossy scroll has done, love his art

[glossy scroll]

Honestly, I gotta say that I am indeed super satisfied with that D&D style GPT. It blows my mind every time it makes a new gen.

[shut niche]

It's making consistently great images. You should be proud.

[glossy scroll]

I'm gonna have a talk with wizards of the coast and see if I can make them pics, lol

Would probably save them a lot of money.

[green light]

Dall-e has lost its mind I asked it to make some Viking sleeve tattoos and it did. Then it refused to do it because of content issues so I asked chatGPT why that is possible and it said because some of the symbols are used by extremist groups which is vaguely true and insane if a small group of idiots supercedes someone's culture and heritage.

[glossy scroll]

Then just ask to make it without the symbols

Or create alternate symbols

[green light]

The symbols are pretty specific. Am I only able to post 1 picture at a time here?

[green light]

That looks amazing.

[glossy scroll]

I like this one a bit better.

[glossy scroll]

You can't have more than 5 images at a time.

[green light]

It's cool how it's roughly the same looking wizard. The 2nd does have a bit more pop or vibe to it. I asked to make a Viking Berserker walking through the fog wielding dual axes in a bear skin and got this two.

[glossy scroll]

Those look cool. Like male models for viking specific commercials

[green light]

Lol I did it with Scottish Highlanders and this dude got a man bun

And a well croft beard.

[glossy scroll]

Hey wait a minute...That viking looks like my non-server profile pic

His face.

[green light]

Which one?

[glossy scroll]

The blue tatted face one

And my main profile pic

[green light]

Oh that's a Scottish Highlander. I have a bit of a wild family history so I was going with various themes all the way to WW2. lol are you serious like it's you or a picture you use?

These are the others it did. I can't even wrap my head around how it can do this.

[glossy scroll]

No it's me

That's how I look like in real life

Amazingly handsome

And dashing

[green light]

I see what you did there 😉

[glossy scroll]

No, but I made that art in midjourney using Hugh Jackman and Jake Gyllenhaal's faces combined.

lol

And aged of course.

To resemble Pythagoras, lol

[shut niche]

Check this out. I told it to re-imagine it as an angry snow beast, battling a wizard. I'm liking the ferocity it's getting in the expressions.

[green light]

Oh wow I hadn't heard of midjourney. So that was in mid journey not Dall-e? Lol that is aged Hugh Jackman and Jake Gyllenhaal. That is wild.

[glossy scroll]

Haha, yeah, I made it on midjourney a few months ago.

[green light]

That is on point.

[glossy scroll]

[green light]

I'll have to check it out. It made these then stopped working.

[glossy scroll]

Those look super cool!

Ted is also big into making tattoo art

You should see some of his, they're wild

Hang on, I think he showed me some, I'll show you his

[green light]

Is he on here?

[glossy scroll]

Yeah, I don't know why it won't let me tag him. But these are his images.

@Ted#2558

I don't know what's going on with the tagging. Not working.

[late blade]

you can't tag a person if you put a hashtag

look for his alias on this server

he's just ted in here apperantly

[glossy scroll]

Yeah, lol, I can't find him on the list when looking or in the search bar...But he's in here.

Somewhere...

[late blade]

if you do search for the server and put from: and the name, you can see his alias in this server and the discord id

so yeah (at)Ted is the right one

[glossy scroll]

It's not working.

I don't know. But it's that guy who posts the female military soldier.

Like in graphic novel style.

[late blade]

ya I know who you mean

[dim cradle]

just sharing cuz i made this a little while ago, similar knotwork

[glossy scroll]

@hot rain

There we go. I remembered saying something to him and just looked up what i said.

[dim cradle]

he's really dedicated

[green light]

It looks Yggdrasil too.

[green light]

Yah it's pretty cool and interesting be was able to show it on a person.

[glossy scroll]

Yeah, he's really good with his artwork. I love the themes he sticks to.

[glossy scroll]

Haha, yeah that would be cool.

But we talk in the DMs, you know, buds

Exchange art ideas

Like I do with Shon and Dys and Aced

[green light]

I'm sure that's pretty helpful with building prompts. It's pretty cool the stuff you all have come up with. It amazes me how Dall-e has progressed.

[glossy scroll]

Yeah it amazes me too. And I can't wait for the next update to see what's up next. So we can hopefully take the prompting business to the next level.

I want to be able to command a little easier.

Right now it's still a bit difficult.

[late blade]

we do, but atm doing xmas stuff, so I'm jumping in and. out of existance a lot today

[glossy scroll]

Yeah, xmas everyone is busy.

Except for me.

[dim cradle]

here's an ornament for your tree

[glossy scroll]

lol, thank you. It's a magnet?

looks like some kind of transducer at the top

[dim cradle]

it's some kind of lifeform

[glossy scroll]

[dim cradle]

gpt-3.5 calls it a "spectral consciousness entity." --"A fusion of technological and ethereal elements, a sentient being comprised of electrified matter and radiant energies. It embodies both AI and spiritual energy, blurring lines between tech and the metaphysical, resulting in a unique and mesmerizing lifeform." So there ya go.

[glossy scroll]

Hmmm. Interesting. Blurring the lines. I like that.

[dim cradle]

Blending and making connections...

[late blade]

it might be me or dall-e, but people use mesmerizing, intricate a lot on their prompts

[grizzled loom]

glass monsters are fun

[glossy scroll]

This is how the revamp came out using your pic

I also used your description for it too.

Second revision

[shut niche]

Yes they are! Looks killer.

[dim cradle]

creative and delicious

colorful, nutritious

[green light]

lol did you do this with Dall-e because that is dope and it did not like my idea of asking it do have batman with David Harbours version of Santa and now it is saying it it's unable to generate pictures like batman because of content policy.

[pulsar sundial]

Hah

[pulsar sundial]

Yes

[dim cradle]

looks very 3D

[glossy scroll]

Snow Beast

[shut niche]

My GPT's ability to work off of images I provide it is really surprising me. This is it's take on your image.

[glossy scroll]

Christmas Dragon

[pulsar sundial]

looks great

[shut niche]

Now that's cool!

[green light]

Have you saw that movie?

[glossy scroll]

Even cooler than this one?

Christmas Half-Dragon

[pulsar sundial]

Which?

[shut niche]

I like that dragon. It's cool with the old bus, and the scale next to the houses and people. I feel like that crate he's carrying should be a boombox, lol. Rock'n out to some metal.

[glossy scroll]

You better watch out, you better not cry
You better not pout, I'm telling you why
Santa Claus is comin' for youuuuuuu

[shut niche]

I gave it that image you made, and then told it to have fun and do its thing. I'm stoked it's working so well just off of images.

[shut niche]

That reminds me of a Berserker I made earlier for my wife. I made Santa and Mrs. Claus. I'll have to find it for you.

[glossy scroll]

Merry Axemess

[shut niche]

I think that's all I asked for to, was just a berserker of Santa and Mrs Claus.

[shut niche]

I like this one better. His battle axe looks great. If you're focusing on these characters, that might not be a bad idea, to specify all weapons should look great. Like with a sheen to them.

[shut niche]

I'm not making any more bird people, lol.

[pulsar sundial]

ahahaha why

[glossy scroll]

I'll try that

Ho Ho Ho!! 🎅

He's using some christmas magic!

[shut niche]

He needs a sleigh that matches his motif. Some raging reindeer.

[green light]

Violent Night with Dennis Harbor as Santa Claus.

[glossy scroll]

You're giving me ideas!

[green light]

Santa just dropped this meat tenderizer off to me.

[shut niche]

@glossy scroll Plug this into your GPT and see what it wants to do with the image.

[late blade]

you got the wrong cat

[glossy scroll]

Merrrrrry Christmas everyone!!

[glossy scroll]

Let's see here...one moment

That ice giant looks epic btw

[shut niche]

Now 'that's' a Santa!! Nice work man.

[pulsar sundial]

did not watch

[shut niche]

I like the ice cavern it created to go with the character.

[green light]

Violent Night is an action comedy with Dennis Harbor where he is a Viking Berserker so it's right up the alley with all these images.

[late blade]

my prompt was:

An image of an immensely big ice giant, holding a large ice spear, and ice armor, walking through a ice valley

[shut niche]

It's cool to see how custom instructions really does give ChatGPT a different "brain" so to speak, in it's tastes and style.

[shut niche]

We weren't prompting, but using vision to copy-expand on the other's image.

[late blade]

ahaaa

[shut niche]

A test of GPT interpretation.

[late blade]

oh nice, I should prob try that

[glossy scroll]

Yeah, testing our GPT's style interpretations

[glossy scroll]

Yeah it's awesome!

[late blade]

that's just api right?

[glossy scroll]

Custom GPTs

Not API

[shut niche]

ChatGPT hiccupped, so I got some "freebie's" after hitting regenerate, and it changed it's mind on the style. I'm liking this battle axe. I love ice and glass in AI art. It's one of the coolest things. I hope they do more with ray tracing types of enhancements in future iterations, to really go bananas with clarity of translucent objects.

[glossy scroll]

Awesome gift btw!!

[grand star]

Hi! I got a time cap of using DALL-E. But I am a professional animator. Can I get more time with Dall-e somehow?

[late blade]

oh I don't see any vision custom api from openai

[shut niche]

We all just wait it out. It's 200imgs per day, with 15/60min thresholds you can hit for rapid fire generating.

[glossy scroll]

Aaaaaaand I capped out

[shut niche]

I probably will in a min b/c I'm running quads.

[glossy scroll]

Time to go be with the fam for a bit

[shut niche]

But it's Christmas, and I'm leaving in a minute, so if I hit it I hit it. LOL

[glossy scroll]

You hit it!

lol

[late blade]

ok, I'm setting up a project for vision with the api

[hot rain]

Telling it a specific style of tattoo goes a very long way. For example, all the gens I've worked on for tattoos have been Tebori tats, and I've made sure to include that when prompting GPT for these.

[shut niche]

That looks great man!

[hot rain]

Thank you!

[shut niche]

Everyone that knows me well is always surprised that I don't have all sorts of tattoos. I think if I did though, that style would be one viable option. 🙂
My issue is, I'm extreme ADHD. I'm know myself well enough to know I would get it, love it, and grow sick of it, and it would still be there. haha

[late blade]

no wonder i didn't find it, dall-e and vision are separate models

[glossy scroll]

I have tattoos. If you get one, make sure its masterclass artwork.

You'll start to regret it soon if its not, lol

[hot rain]

I'd love to get a tattoo like this personally, but it's a tremendous time investment and isn't cheap either, and artists for this specific tattoo style are far and few between in the Western hemisphere where I live.

As with a lot of what I make, this is for the main character of my book series. Long story short, she grows up on a small fleet of ships that function as their own mobile safe zone during a zombie apocalypse. The fleet rescues people from all over the world over the years, and thus has an exceptionally diverse demographics makeup. One of her role models is a Kenjutsu master rescued from Japan, and after his tremendous impact on her life she eventually seeks out a tebori tattoo once she has the means to do so later in her life. So it's sort of a personal tribute to her old sensei, in a way.

[thick smelt]

NOOO NOT santa being jacked! 😢

[chilly onyx]

Competing with @empty kelp

[glossy scroll]

You wish you got the right one for christmas, lol

[thick smelt]

no i made it for laughs and giggles

[shut niche]

Are you Japanese? Or just into the culture? Sounds cool btw, thanks for sharing. It's cool to see the different ways people are using this tech.

[chilly onyx]

Oh, no I meant that Santa will be competing with the one from @empty kelp gens

I see a Santa battle approaching

[glossy scroll]

Haha. Somehow, all women seem to like that jacked Santa look...I wonder why

[green light]

What sort of prompt did you have to use to put it on a person because it would be good to do that with the images I posted above because I plan to turn them into tattoos.

[glossy scroll]

it's a secret

[hot rain]

Just into the culture partially, but I learned a lot about it doing research for the Kenjutsu master I mentioned for accuracies sake.

[shut niche]

This came out pretty nice. A few of us have been refining GPT's. Mine is really surprising me with it's use of vision in, to create image out. So this is based on your image from above.

[hot rain]

"Make an image of a blonde caucasian female character with a tebori back tattoo. Make in the style of a graphic novel or comic book. The woman is a survivor of the apocalypse and should have slightly toned muscles."

That's roughly the prompt I've used.

[shut niche]

Oh, and I told it I wanted it to be an oil painting, just to mix it up a little.

[thick smelt]

Luigi holding a bucket of kfc because
It started in the early 1970s, when KFC was still new in Japan. It began marketing chicken as the country's Christmas meat with the catchphrase “Christmas is Kentucky.” Every Christmas Day, people queue down the street outside KFCs like this one to pick up their Christmas packs.

tldr: Japan had chicken for christmas 🐔 🌲

[hot rain]

Another run of my prompt:

[shut niche]

It did one of a guy too. I guess it couldn't tell the gender. Whoa, you just shared that as I was trying to show you this. Uncanny how similar they are.

[green light]

lol what prompt did you use because I wrote Luigi holding a bucket of KFC to see since it's been citing it doesn't follow the content policy for other things that are copywrite related.

[glossy scroll]

Yeah, it's interesting how Japan took that KFC thing to the next level, lol

[thick smelt]

reprompt because luigi had fries on big bruh

[thick smelt]

i used bing

[hot rain]

It does default to male like, 30% of the time. It's frustrating, but probably to do with the censors clashing with the concept of a topless female even if it's just her back.

[glossy scroll]

@green light You can use product images. Product and franchizes find it to be advertising.

It's not against the copyright restrictions.

It's promoted, not restricted.

And somehow, the brands all look perfect every time.

[green light]

Thank you. That worked. Does Bing not use Dall-e?

[hot rain]

Extreme amounts of the exact same images in DALL.E datasets is the most likely reason that the logos all come out so cleanly. It has millions of examples of each to draw from.

[green light]

Within chatGPT? So using the prompt "Luigi holding a bucket of KFC in Japan" should have worked?

[thick smelt]

they do, dall-e 3

[glossy scroll]

Yeah, it should work.

[green light]

Weird. I got this.

[glossy scroll]

[green light]

It keeps doing this to me today though. It wasn't at all last night.

[glossy scroll]

That's strange.

[glossy scroll]

Does it?

[green light]

My grandpa was in a group called the ladies from hell in WW2 so it made this yesterday but won't today.

[hot rain]

In a similar vain, I got it to make this Perfect Cell art for a video I made, but it has refused to ever make another one of him for me due to policy restrictions.

[glossy scroll]

Epic Ted!

Epiccccc!!!

[glossy scroll]

lol, that's a funny image of that guy playing straight into battle

[green light]

So it worked once then wouldn't after that? It's so weird. I even tried to show it it made those image's and it said it can't comment on that. It's like it's being passive aggressive.

[open trench]

Wow

[thick smelt]

some christmas game covers made by ai

[glossy scroll]

This is honestly not that bad, but some minor tweaking would make it perfect! Keep tweaking!

[glossy scroll]

Oh, that's actually pretty good.

Surprising how clean it looks.

[thick smelt]

some pinoy christmas tradition as games

[thick smelt]

ty ! ^-^

[green light]

That's what they did. They'd play the pipes and march across the field and when they Axis heard them coming they'd flee because they knew what was coming for them. Hence the name ladies from hell. This pic is of a real one.

[chilly onyx]

Here’s one of mine

Took me a long time since the folks don’t exactly prioritise the internet here…

[thick smelt]

pinoy stuff as ai art cover for games

[shut niche]

I did a double take. Nice work man! Looks cool. I'm digging the dystopian wreckage under the snow.

[hot rain]

Last night it gave me a few "Christmas Truce" gens before deciding policy restricts it from making art of real events.

[thick smelt]

the yeti would give thor a run on his money 🤣, like the weapon, more cooler than mjolnir!

[green light]

That looks dope but it's so weird. Bing isn't restricting me. I asked it for specific symbols and made this. Thanks for the pro tip @thick smelt

[thick smelt]

salamat! (thank you in filipino)

[green light]

Mabuhay!

[hot rain]

I wasn't trying to get Storm Trooper armor, but this is literally an amazing design for a heavier Storm Trooper lmao.

[thick smelt]

another filipino christmas food made by ai (since the name of the food is block by ai mods i put purple tube shaped rice cake instead)

[thick smelt]

here's the original photo from google for comparison

[thick smelt]

which one of them is ai?
🅰️ or🅱️ ?

[late blade]

@formal osprey I tried the JSON thing again, it works perfectly if you upload a JSON file to dall-e instead of using the prompt with the notation

[late blade]

just tried an old photo of me with vision and then ran it through dall-e, and this is what I look like apperantly

[green light]

Where in the Western hemisphere are you? There are some in Toronto.

[dim cradle]

tattoos of some game you like in 20s or 30s seems risky long term haha. imagine bein 64 and like oh yeah this big ink block on my back well it was a video game 40 years ago you see and now you branded yourself for life with it

random christmas thought i guess lol

[green light]

That's why it's best to know thyself and pick something that resonates with you and your identity. If that changes then the person is a personality disorder. So there is that.

[dim cradle]

well thats my point most people mature always. getting a big video game back seem like a good idea when youre young but...

youre prob not gonna be all about Shinobi 19 when youre in your 50s or 60s

[dim cradle]

what do yiou say just remake this image in dalle?

[green light]

I did not prove your point.

[dim cradle]

thanks for trying at least

[late blade]

just used GPT4V, then make a JSON file and then feed it to dall-e and make it in that art style

[thick smelt]

That looks cute, like a found lost media for an old anime from the 90s!

[late blade]

yeah lol

thing is, if you put the original photo, it's pretty accurate

[dim cradle]

i dont even know what a json is lol 😕

[late blade]

just be careful when doing this kind of photo, I allowed me to do that because it's a photo of me, but do be careful when doing other persons

[green light]

That is not what happened. I proved your overgeneralization is wrong. So there is that.

[late blade]

you could use vision and let it describe the photo and then put it in dall-e

[dim cradle]

whoa. that one is epic

take THAT, mj 6.0 haha

[thick smelt]

Wow your "doom at home" art is fire 🔥

*doom the video game

[glossy scroll]

Tsank you guys!

[thick smelt]

Yw

@glossy scroll you dropped this → 👑

[dim cradle]

I wish it had that thing... I forget the name of the AI now ow... the one you can do at home if you have a big video card, but you can do image 2 image in it. imagine that in Dalle 3 wow

stable diffusions

[thick smelt]

Big Pytha and dalle 3 W

[dim cradle]

is it with a gpt you made ?

[glossy scroll]

Yep

It's the same one I use to make DnD pics

[dim cradle]

nice

[thick smelt]

The art looks more like a doom concept art to me 😅 (or Metroid)

[dim cradle]

i thought maybe master chef haha

[glossy scroll]

I got a Doom for ya

[thick smelt]

Yeah that's cool

[glossy scroll]

I got your Master Chief for ya

Maybe a Spartan war?

[thick smelt]

Here's some animal crossing too

[glossy scroll]

That was a shift of genres, lol

But still lovely

[thick smelt]

Imagine master chief having a restaurant named "MasterChef"

[thick smelt]

Ty

[glossy scroll]

Didn't you know the real face of Spartan-117? It's Gordan Ramsey

See, I told you.

[thick smelt]

The pun
MasterCHEF 🤣

[dim cradle]

what do the expert reccommend for making an image gpt... just put in the instruction all the photo or image detail you want to always use, then when you make the prompt just write the subject because the gpt will do the rest?

[glossy scroll]

Follow this guide I made, explains everything #1187233013956874260

[dense mesa]

Function createCombinedImage(descriptions, styleAttributes)
    /*
    Takes multiple character descriptions and style attributes and combines them into a single scene.
    */

    // Define the base scene
    baseScene = defineBaseScene()

    // Add characters to base scene
    For each description in descriptions:
        character = createCharacterFromDescription(description)
        baseScene = addCharacterToScene(baseScene, character)

    // Apply style attributes to the scene
    For each attribute in styleAttributes:
        baseScene = applyStyleAttribute(baseScene, attribute)

    // Generate the final image
    finalImage = renderScene(baseScene)

    return finalImage

Function defineBaseScene()
    /*
    Defines base parameters.
    */
    return new Scene(background="space battle", ambiance="whimsical yet intense")

Function createCharacterFromDescription(description)
    /*
    Creates character image based on description.
    */
    character = ImageGenerator.generate(description)
    return character

Function addCharacterToScene(scene, character)
    /*
    Adds the generated character to the scene.
    */
    scene.addCharacter(character)
    return scene

Function applyStyleAttribute(scene, attribute)
    /*
    Applies given style attribute to the scene.
    */
    scene.applyAttribute(attribute)
    return scene

Function renderScene(scene)
    /*
    Renders the final scene.
    */
    return ImageRenderer.render(scene)

// Character descriptions
descriptions = [
    "cheerful animated cow character in a sweater with a backpack",
    "charming dinosaur creature with a bowtie and a suitcase",
    "futuristic armored soldier with a visor against a space backdrop",
    "space marine in power armor with a weapon surrounded by a fiery glow"
]

// Style attributes for combined scene
styleAttributes = ["cuteness", "whimsy", "high-action", "sci-fi"]

// Create combined image
combinedImage = createCombinedImage(descriptions, styleAttributes)

[glossy scroll]

There's a lot of unnecessary instructions you have in there. You could make a lot better image if you just toned down the punctuations a bit. And I promise you that putting words next to each other don't do anything special. That's just my opinion.

[dense mesa]

I didn't write that

I wrote a gpt that wrote that

that was the first iteration

[dim cradle]

oh sure, blame it on the AI!

[glossy scroll]

@dense mesa I do like your goals though. If that is a GPT you made, I think it's awesome.

[dense mesa]

combo on the last one

[glossy scroll]

[dense mesa]

here's what it did there

Function reimagineInStyles(baseImage, styles)
    /*
    This function takes a base image and reimagines it in four distinct styles.
    */

    // Initialize an empty list to store the reimagined images
    reimaginedImages = []

    // Loop through each style
    For each style in styles:
        // Reimagine the base image in the current style
        reimaginedImage = applyStyle(baseImage, style)

        // Add the reimagined image to the list
        reimaginedImages.append(reimaginedImage)

    return reimaginedImages

Function applyStyle(baseImage, style)
    /*
    Applies the given style to the base image.
    */
    // Determine the style and apply it
    Switch (style):
        Case 'voxel':
            return convertToVoxelStyle(baseImage)
        Case 'watercolor':
            return convertToWatercolorStyle(baseImage)
        Case 'pixelArt':
            return convertToPixelArtStyle(baseImage)
        Case 'oilPainting':
            return convertToOilPaintingStyle(baseImage)
        // Additional styles can be added

// Define the base image and styles
baseImage = loadImage("base_scene_image.jpg")
styles = ["voxel", "watercolor", "pixelArt", "oilPainting"]

// Reimagine the base image in the four styles
reimaginedImages = reimagineInStyles(baseImage, styles)

[exotic lark]

Hii

[chilly onyx]

The spotty internet connection really slowed me down

Hmm… missed the decorations theme. Have been half working on a GPT that creates photos of small decorative home sculptures

[glossy scroll]

Would love to have that for my computer table!

[dim cradle]

in the penalty box already 😭

oh well, Bing is King

[late blade]

what did you do?

[dim cradle]

i just mean used all the prompts i guess

i only get 25 these last days it seem

some cool stuffs in daily theme. i never really look there before

lot of names i do Not recognize... we must scare them from discussion haha 😂

[late blade]

we are over 18k people online atm, sure a few are there

[dim cradle]

i see catgirl in hall of fame and i know its dys topia work haha

[late blade]

lol

[glossy scroll]

I'm only gonna put my worse stuff in daily from now on

[late blade]

I'm just putting some stuff while I experiment

[dim cradle]

i wish i have the money I spent on MJ 6.0 to use on api for dalle. 😭

[late blade]

just enjoy what you can do,

and be creative

[dim cradle]

i am happy with Bing really, i love the variants and it always make 4 image. i just wish it had a longer text entry

[glossy scroll]

I just said give me a random kung fu martial artist and it used Hiroyuki Sanada for its face, lol

I mean he is kind of legendary though

So I don't blame Dall-E

[late blade]

I'm so bad with actors names and faces, you could put a cat in front of it and tell me he's the rock and I'd beleive it

[glossy scroll]

Lol

I've just seen so many kung fu movies as a kid, I got these faces vividly trapped in my psyche.

[late blade]

hehe nice

[chilly onyx]

Same, no idea with faces 😅

What were you going for?

[late blade]

something more around joy and hearts

[chilly onyx]

Not really close to what I was after though tbh 😆

Asking to redo it with a Warner pallets and human emotion and see what happens?

Curious

Plus I’m enjoying Dall-E semi-vicariously right now 🙄

[late blade]

I'm going away from swirls and twirls and spins and what not. too many of those in the past days and the ornaments crowned that

[chilly onyx]

It does enjoy a well crafted swirl

[late blade]

o.o

so storm troopers

[dense mesa]

Imagine a scene where a heroic space warrior stands guard at the entrance to a submerged temple, echoing the deep-sea explorer's discovery, representing the warrior's role as a protector of history's mysteries.

[late blade]

interesting

[dim cradle]

nice

i get it to make some Batmans once by i thikn calling him a super hero bat-vigilante haha

ofc in Bing (is King) you can just say Batman 🦇

[dim cradle]

noice

you can tell when a model was train by Batman... in Bing he is Afflecks in MJ 6.0 you get robert pattson

[dense mesa]

lol

[dim cradle]

one thing that is impressive in Bing... if you change the year say "1988 Cinematic Image of Batman" or "1971 cinematic image of Batman" it will modify his costume to the times. 80s will have more Mike Keaton touch, earlier the old tv show, 2000s you get the Nolan guy, more modern Afflecks and some merge all together very cool

👑 you drop this, Bing

[glossy scroll]

Welcome to Dall-E discussion!

[late blade]

winnie the pooh is public domain, so that's not a problem, but the artwork done by disney could be a problem

[dim cradle]

awesome

[late blade]

what's a long screenshot?

[hot rain]

Stormtrooper Astartes:

[boreal gate]

Some web browsers offer it. It's like taking a screenshot while continously scrolling, making it a longer image with more information

[late blade]

there are only wide, portrait or square formats provided by dall-e 3

[boreal gate]

This was the output

[late blade]

portrait format

[boreal gate]

Yea

[late blade]

then use portrait, long screenshot doesn't say much to dall-e

[boreal gate]

Oh well that was something of what I was looking for, I wanted to see how long it can make the images

[late blade]

there are fix resolutions, 1024x1024, 1792x1024 or 1024x1792

[hot rain]

I've tried getting a minigun before, this is the first good one I've gotten out of AI.

[dim cradle]

that looks like an AI you better not mess with haha

[hot rain]

"Create an image of a unique apex predator."

[dim cradle]

epic

[glossy scroll]

2D arts rule!

[hot rain]

"Name: Goliathorn, The Shadowed Fury

Stat Sheet:

• Classification: Suprapredator Strain XX

• Size:

  • Length: 50 feet from snout to tail
  • Height: 22 feet at the arch of the back
  • Weight: 15 tons

• Physical Attributes:

  • Skin: A composite dermal layer of interlocking scales and chitinous armor
  • Claws: Hyper-extended, razor-sharp, capable of rending steel
  • Teeth: Multiple rows of serrated teeth, able to regenerate
  • Tail: Long, prehensile, with barbed spikes
  • Eyes: Four, with 360-degree vision and night vision capabilities

• Sensory Capabilities:

  • Olfactory: Advanced olfactory receptors for tracking and hunting
  • Auditory: Capable of detecting infrasound for communication and hunting
  • Visual: Superior depth perception and thermal detection

• Defensive Mechanisms:

  • Armor: Natural body armor impervious to small arms fire
  • Regeneration: Rapid cellular regeneration from injuries

• Offensive Capabilities:

  • Bite: Over 16,000 pounds per square inch bite force
  • Strength: Able to topple structures and uproot trees
  • Speed: Capable of bursts of speed up to 40 mph
  • Stamina: Endurance hunter, can pursue prey over vast distances

• Behavioral Traits:

  • Dominance: Assertive and aggressive territorial control
  • Intelligence: High problem-solving skills, memory akin to an elephant
  • Hunting: Stealth and ambush predator, with tactical planning abilities

• Diet:

  • Hypercarnivorous: Consumes large mammals, other predators, and is known to attack anything that enters its territory

• Habitat:

  • Prefers dense, dark forests or jungles that provide cover for ambushing prey

• Reproduction:

  • Reproductive Method: Oviparous, laying clutches of eggs in secluded, fortified nests
  • Clutch Size: Typically 2-4 eggs
  • Maturity: Offspring reach full maturity within 5 years

• Lifespan: Estimated at up to 150 years in the wild

• Conservation Status: Not applicable (species not recognized by conventional taxonomy)

Goliathorn, The Shadowed Fury, is an apex predator designed for dominance and survival. With its formidable combination of physical power, sensory acuity, and natural defenses, it stands uncontested in its ecological niche. The presence of such a creature in any habitat would dramatically reshape the ecosystem, placing it firmly at the pinnacle of the food chain."

[dim cradle]

nice

[glossy scroll]

Amazing. You've done it.

You've made your own wicked art style!

[late blade]

hmmm, seems dall-e got moody with me, and I'm now in the penalty box

[dim cradle]

ted ftw

sorry to hear that dys topia

[late blade]

lol, that happens

it's been kinda weird forgetting stuff

[green pebble]

I gave the prompt a shot

[dim cradle]

awesome

just has one too many hind leg haha. then again... 🤔 hes a monster who can say how many he has lol

[green pebble]

Exactly lol

[weak hull]

I am trying to build a prompt or custom gpt that will give me a 3D render of a module that is always roughly the same size and framed the same way each time. I want the module to be 1 unit high, 2 units long, and 1.5 units wide. Mainly I want it to be something other than a cube which is what I keep getting. Any advice would be greatly appreciated. I'd also like to know if this is an impossible ask before I pull out what little hair I have left.

[late blade]

hmmm, i wouldn't know how to do that tbh

@weak hull maybe add:

...ensure that the shape is unique and distinct from a cube. Please provide variations in the shape of the module.

[chilly onyx]

I did some fun Orks around the Halloween pumpkin theme. The main challenge I found with Orks was for them to not resemble Spare Marines

I also had to check Ork lore regards the beard

It’s a Squig. Were all good 🙂

[dim cradle]

bad ass!

[dim cradle]

what are the tubes to do? charge her up?

[late blade]

she's tanking energy

robots don't eat normal food lol

[dim cradle]

shes at the tesla station

[late blade]

lol

dall-e keeps forgetting stuff >.<

i still need some tweaking but I'm getting there

need to think of another way of prompting while using json files, seems chat is more forgettful that way

[late blade]

well seems I'm finally getting full body images, but still not in the portrait format. 😠

and a penalty box again until 1:24 AM, seems it's time for me to go to bed

[dim cradle]

[vapid elk]

that first one if super cool

[chilly onyx]

Oh, yeah I only included the two Battlewagons to show how much it wanted to turn them into Space Marines. I liked the way it rendered that first one though as well 🙂

[dim cradle]

The extra body parts makes photos interesting

[empty kelp]

#daily-theme message
Attempting to create wildlife photos in DALL-E. The key elements in the prompt were, "This is very serious." and, "The image should be really, really amazing."

[dim cradle]

good quality

short rabbit

[empty kelp]

this is good, but i'm not sure the robot AI should be putting broccoli in the stir fry

[dense mesa]

he's still learning

[dim cradle]

he makes great panang curry and bubble tea though

i know its dys topia image without seeing the name when i see a catgirl and pictures when i see a food table haha

[dense mesa]

I just weave the same concepts in over and over to see what happens

I do like this style

the image thread is up to 22144 tokens so might have to move on soon

[dim cradle]

what is max 25k

[dense mesa]

I don't know. it just starts to kind of get bogged down after a while. plus doesn't remember most of the conversation

[glossy scroll]

It mostly has to do with overworking your commands and instructions.

#1187233013956874260

[dense mesa]

but keep mixing in new things

these instructions are pretty on point when I use them along with my general approach to things

[dim cradle]

nice

[dense mesa]

I don't know. my approach is, they might not all be winners, but I can always reintroduce the concepts later

[dim cradle]

it knows the power armor

[green trout]

how would you prompt it to get a panorama files for a 3rd party stitcher?

I'm still fascinated with paleofuturistic Egyptian Battle-Mechs myself

[dim cradle]

mech warriors are fun

[dim cradle]

it's harder to depict a human operator within

[empty kelp]

Can you please draw a wildlife photo of a gecko on a lava rock. The gecko is the best of all possible geckos. In the entire history of the universe nobody has ever seen a better gecko than this. The image should have 1024x1792 resolution, portrait orientation, and the best possible HD rendering.

[dim cradle]

"please" 🙂

[dim cradle]

so vivid and vibrant

[empty kelp]

Having good results with putting this at the end of a ChatGPT Plus request:
The image should have 1024x1792 resolution, portrait orientation, and the best possible HD rendering.

[dim cradle]

thats awesome shon

[empty kelp]

Try telling the AI to use, "The best possible HD rendering"

[dense mesa]

tell it to use lanczos from the PIL library to upscale your image

[empty kelp]

i've been using Gigapixel AI and Photoshop to upscale things

[dense mesa]

fair enough

[empty kelp]

will investigate lanczos

[dim cradle]

anyone ever tried Magnific?

to upscale

still perfecting my concept of the joyful meditating man

[dim cradle]

i'm skeptical of upscaling lol

[dense mesa]

well there are lots of different types

[dim cradle]

me too from what i saw of som emagnific

look to change the image too much but i do not use it

[empty kelp]

nothing i've posted here has been upscaled or edited. it's all straight from DALL-E

[dense mesa]

alright, so there are algorithmic primitive upscalers like lanczos that I mentioned earlier

but then there AI upscalers

there aren't really any one size fits all upscalers because it very much depends on what you're going for and what you're upscaling. different approaches that bring out different things

[dim cradle]

i think i'll just wait until the image generators natively support higher res, probably won't be long

[dense mesa]

either they retrain at a higher resolution or they upscale

[dim cradle]

i think they train high res images at low res because it takes less processing--they can generate at arbitrarily higher resolutions but are fixed at lower resolutions now due to capacity and demand, at least that's my impression.

as to the tech details, that seems to be a blackbox for now in general

[dense mesa]

really can't though because the model splits the space up into pieces. if you add pieces it doesn't know how the pieces stitch together. so lots of errors

need to double check, but 16x16 or 32x32 pixel squares I believe

seems like 32x32, but could be off on that

[dim cradle]

well, at least it's better at hands now. i still see a third leg now and then.

[dense mesa]

yeah, so if you render at a resolution higher than it's training you'll end up with lots of extra limbs

and missing limbs

[empty kelp]

if you say, the best possible HD rendering with ChatGPT Plus it seems to do unusually good lighting with light beams and subsurface scattering type effects -- But it also seriously limits the number of images you can create before a timeout

[dim cradle]

my meditator is not abstract enough

hmm, maybe a character

let's add some body art

man i wish bing used the hd setting on the api

[fierce raven]

wassup

dall e is cool

but my credits ended

[empty kelp]

yo

[fierce raven]

how long does the chatgpt api last if i use in a app?

for image gen

dall e

[formal osprey]

I just tested again and same result

can you verify the prompt chatGPT used to generate your picture?

[formal osprey]

you can check with this button

[fierce raven]

just generated a pfp

[dim cradle]

Poseidon inspired?

[glossy scroll]

Because of the Triton? I didn't put that, but it just came out like that. Was just trying to make a random sea species.

Flying Turtle Power!

[dim cradle]

Not something you see everyday

[glossy scroll]

Hehe

[glossy scroll]

Very nice!!

[dim cradle]

That’s remarkable

The hair is perfect

[glossy scroll]

I have a sea species too, from before

They kinda look similar

[dim cradle]

he has great hair also

[dim cradle]

very lifelike

[dim cradle]

haha it's got a smile

[dense mesa]

good vibes there

started a new image conversation. sometimes takes a few rounds to get it on my wavelength

[dim cradle]

mmmhmmmmmmmmm

it's fun to generate a scene, then request another reimagined for another century.

[dense mesa]

double neck?

[dim cradle]

you get interesting results when you paste in a long art prompt you've crafted with gpt-3.5 into one of our fancy custom gpts

[dense mesa]

yeah, I have sorts of different approaches. they do come together sometimes

[neat condor]

Call

Call

[dense mesa]

Call

[neat condor]

Who are you.

[glossy scroll]

He uses Rocky Mountain forest mint shampoo

That's why his hair is so flowy

[serene umbra]

wow it's cool.

what are you think of this photo.

[dense mesa]

that's SDXL

[glossy scroll]

Fake

Herbert and I play this game everyday, lol, he asks me what's real or fake and I have to guess 🤣

It's fun, I like it

[serene umbra]

haha yes

I'm glad too

[dense mesa]

ahh, lol

fair enough

[glossy scroll]

[dim cradle]

xenomorph blended with a praying mantis

[dense mesa]

mix this in

[low dew]

Add this before the prompt you used to get that to mix it in. Share results! Can’t wait to see!

[dense mesa]

well I have to wait 35 minutes

[low dew]

[dense mesa]

but I have lizards

[low dew]

So inconspicuous lol

[vapid elk]

nice limzards

[dim cradle]

xenomorph blended with scorpion turned out well on the second try. first try the tail was blended with the head, a most unnatural thing.

the dalle look less photo real to me often but it must have more dpi or something idk they always look better Fidelity than Bing (which is still King)

looking a lot like pro nature shots

[dense mesa]

kind of reminded me of this image

[dim cradle]

impressive amount of different dishes

that must be the boss on the cake haha

[dense mesa]

ample popcorn

[dim cradle]

at least there's some nutritious fruit

[dense mesa]

here's a vintage layout

once again a sufficient amount of popcorn

[dim cradle]

make some food only AI can do. Or absurdly large hamburgers or something haha

[dense mesa]

not quite sure what to think of the latest one

definitely an interesting view

[dim cradle]

looks like it didn't quite finish... a new kind of artwork.

[dense mesa]

sometimes does that when it tries to mix a bunch of things for whatever reason. there's definitely coherence to the different panels

[dim cradle]

yeah, like it accessed the assets but didn't finish blending

[dense mesa]

I think it's a calculated. I think there's actually a name for that approach. but definite malicious compliance vibes

[low dew]

Bro lmao that’s the last supper

[dense mesa]

seems like a proper AI fever dream meal

[low dew]

Lmao

[empty kelp]

Can you please draw a wildlife photo with several diverse geckos crawling all over a pile of large lava rocks. In the slightly blurry background there are coconut trees blowing in the wind, Mauna Loa is having a spectacular eruption and lava is flowing down into the foreground, and there is an epic rainstorm with lightning and a rainbow. The geckos are the best of all possible geckos. In the entire history of the universe nobody has ever seen better geckos than these geckos. The image should have 1792x1024 resolution, landscape orientation, and the best possible HD rendering. The geckos are wet from the rain. People will start to cry when they see this image because it's so beautiful.

[thick smelt]

umm... guys is this AQUAMAN? 🤣, jk

[dense mesa]

I am very intrigued by these results

compare the details

[neat condor]

I am very intrigued by these results

[empty kelp]

To the top two pairs of images you should try adding this:
Please add the style of Hieronymus Bosch to the image with elements of heaven and hell.

It will make them really go crazy

[dense mesa]

extra shiny skin

[empty kelp]

the clouds, ocean, cliffs, and grass are spectacular. the horses are a bit reflective

[hot rain]

I always liked this concept, so figured I'd prompt it.

"A giant ancient tortoise with an ecosystem on its shell."

Antoids, a race of subterranean savage half-ant monstrosities with an innate hatred for surface dwellers in my friends fantasy world.

[dim cradle]

i prefer the centaur look

[hot rain]

Then there's always the Antoid mages.

[thick smelt]

some game characters doing christmassy stuff made by bing

[dim cradle]

the joyful man is still running

so the legend goes

[bleak mantle]

Merry late Christmas!!

[hot rain]

A bit off topic, but here's the thread of that happening lol. Ridiculous rolls. Open in browser and zoom to read clearly

But more on topic, I really like this one because of the person walking on the beach for additional scale.

And a swap from tortoise and village to snapping turtle and fortress.

[dim cradle]

it works in so many ways

[hot rain]

Right? I hope other people jump in on it with their own styles too 🙂

I desperately want the concept incorporated into a 5e campaign.

And put to the test in battle.

[empty kelp]

trying to get the geckos to aim a bow and arrow

Aloha! Can you please draw a wildlife photo of a gecko standing on a lava rock and carefully aiming a bow and arrow with its two front feet. In the slightly blurry background there are coconut trees blowing in the powerful wind, a massive volcano is having a powerful eruption, and lava from the volcano is flowing all the way into the foreground. There is a heavy rainstorm with lightning and a rainbow. The gecko is the best of all possible geckos. The image should have 1024x1792 resolution, portrait orientation, and the best possible HD rendering.

[hot rain]

Very nice!
I had a leopard gecko, sweet little pets.

[empty kelp]

just added the description (above) if you would like to try it

[hot rain]

Out of generations, but it's on my agenda tomorrow when they're back!

[empty kelp]

for some reason it's easier to get an elephant to shoot a bow and arrow correctly than it is with a gecko

[hot rain]

Maybe "gecko man"?

Adding gecko man might help the AI give it more human actions like pulling a bow.

Like lizardmen from Warhammer

[empty kelp]

this was my first attempt at getting a DALL-E elephant to aim a bow and arrow

[hot rain]

Pretty exceptional lol

[empty kelp]

the geckos take a few attempts to get the bow and arrow to draw correctly, but they have kind of amazing foot placement. their back feet grip onto any weird angle while they're aiming

[thick smelt]

this is kinda canon Wait a sec, why Mugman has red nose?, i made it with bing

[hybrid fox]

I love that there is another fellow cuphead enjoyer here

[chilly onyx]

Dunno about the one with Big Ben man, that looks like there has been a pretty horrific automobile incident 😅

[thick smelt]

i did it, the ultimate Nintendo and Hasbro thing in bing, mr. potato head dressed up as Luigi.

[chilly onyx]

I’m getting closer with my stupid ornament GPT but it’s very hard to avoid it adding in elements that don’t make sense such as floating fragments

It’s been a fun low energy exercise as I’ve been bumbling around the parents house over Christmas looking at their ornaments. I’ve been trying to get it to reproduce these things, just as an exercise in prompting snd working with different materials

[glossy scroll]

Yes that is a problem. But this is a nice one 👍

[glossy scroll]

Do you ask the AI to explain the images to you when you give it a reference image?

Its better for you knowledge so that you can better define the textures of the models.

[thick smelt]

hi pytha!

[glossy scroll]

Hi lugwig

[chilly onyx]

Yeah, tried lots of things. I’ve tried a good few variations now, sortof happy where I’ve landed. It was just a practice project anyhow, not the sort of image generator I’m likely to get a huge amount of use out of

[glossy scroll]

Try using this GPT. It's made by the OpenAI people. And it can help better define your image.
https://chat.openai.com/g/g-fTA4FQ7wj-hot-mods

@thick smelt What are you up to?

[chilly onyx]

It did this to the Robin I posted a couple messages up

[glossy scroll]

Ohhh. I guess it didn't help huh?

[chilly onyx]

Well it’s nice and all, but I like the fractured, missing elements. To me the sculptures are speaking to a loss of something or other. This looks a bit too polished and shiny

Interesting GPT though. I haven’t messed much with OpenAI’s own offerings

[glossy scroll]

OpenAI has got some good stuff

[thick smelt]

doing stuff i did post some ai stuff while you're away tho

here in discord

[glossy scroll]

Really? What kind of AI stuff did you post?

👀

Good stuff I hope

[late blade]

heyo everyone

[chilly onyx]

Hiya

[thick smelt]

YEP good stuff

[glossy scroll]

Hii Dys

[thick smelt]

like giving the blue pikmin the relaxation it deserves after being repeatedly doing the same thing in the game covers, being crushed etc.

[empty kelp]

Aloha! Can you please draw a photo of a gecko with interesting color standing upright on a lava rock on it's hind legs and aiming a sparkling jewel encrusted bow and intensely glowing red fire arrow with its two front feet. The bow's string glows bright white. In the slightly blurry background there are coconut trees blowing in the powerful wind, a massive volcano is erupting explosively, and lava from the volcano is flowing all the way into the foreground. There is an intense storm with heavy rain, lightning, and a rainbow. The gecko is the best of all possible geckos. The image should have 1024x1792 resolution, portrait orientation, and the best possible HD rendering.

good morning!

[glossy scroll]

Hehe. Those pikman never want to listen do they?

[thick smelt]

yeah

[glossy scroll]

Always just giving the resting face

[thick smelt]

there it is blue pikmin relaxing

[glossy scroll]

Lol

[chilly onyx]

Happy toad

[empty kelp]

if anyone wants to try having a gecko stand upside down on a cave ceiling while aiming the bow and arrow it would be a good exercise. i want to try that, but i need to go sleep in a few mins

geckos have powerful suction on their feet, and the feet grip onto things at any angle

The AI seems to have a deep understanding of gecko feet

[thick smelt]

this is what i think of a happy toad 😆 :

[glossy scroll]

Lol

That's a mushroom

[empty kelp]

next project will be to have a creepy spider wield a sword and assault rifle

[glossy scroll]

This is what I think when I look at that

[empty kelp]

did a bunch with a kangaroo holding a two handed battle axe

[thick smelt]

Ikr

[thick smelt]

Because when I see the word toad, I think of the Nintendo character

[empty kelp]

that's a fearsome mushroom stick

[glossy scroll]

But what are the frogs called in Mario?

Are they called Spinach?

This is most unusual.

[dim cradle]

who can buy cradits

[glossy scroll]

You can

[dim cradle]

no

i dont have paypal our card i living in iran

[glossy scroll]

Are you enjoying gormesabzi there?

[dim cradle]

yes

[glossy scroll]

That's my favorite

[late blade]

I can buy credit, but I don't buy credits for strangers.

[dim cradle]

oh you ate ghorme sabzi its very good men

[dim cradle]

its ok

[glossy scroll]

Oh I love it, and some koobideh to go with that.

Mmmmm

I'm getting hungry now

Iranian cuisine is one of my favorites.

[dim cradle]

ok go ate koobideh

[glossy scroll]

Lol. I will

[dim cradle]

me.to

[glossy scroll]

Well yeah because you live there, lol

[dim cradle]

xd

How much cerdits do you have inside the dall-3 @glossy scroll

[glossy scroll]

I don't use credits. I have ChatGPT plus membership.

[late blade]

I will say this only once, this is not allowed by terms of use. Do not ask others to buy credits for you. @shadow bane

[dim cradle]

ok

[dim cradle]

oh ok

its good

[vapid elk]

dalle-3 isn't available on the labs

yo ucan buy credits on the labs, but it is only for dalle2

[dim cradle]

ok

[vapid elk]

dalle3 can only be used via ChatGPT and the API

i mean, you can buy API credits and use dalle3 via API

but labs credits will only grant you dalle2 usage

[dim cradle]

you can accept my friend request ?

[glossy scroll]

Sure

[dim cradle]

Bing is King Daniel

if you cannot use dalle via gpt

go to Bing Image Creator

its also Dalle

[vapid elk]

it isn't the exact same thing, but still really good

[thick smelt]

facts

[vapid elk]

some would say it is better due to the fact it has a weirder auto moderation process

[late blade]

it's dall-e but not dall-e

[dim cradle]

I always have it open and take my gpt prompt for variants. if it would allow longer prompts, it would not just be King but Emperor haha

[late blade]

I started generating JSON files for every image with prompts and metadata

[dim cradle]

for what purpose? 🤔

[late blade]

future references

that way I don't have to look for boundless text

and no, exporting your chats to PDF is a horrible idea

[dim cradle]

capped already 😭

[desert scroll]

You okay bud?

https://th.bing.com/th/id/OIG.tzv765aQLBkntR_W5zJ2?pid=ImgGnhttps://th.bing.com/th/id/OIG.tzv765aQLBkntR_W5zJ2?pid=ImgGn

Generated by Bing Chat AI image generation

[late blade]

Penalty Box for you!!

we have to convince the chat mods and higher ups to add an emoji with penalty box

just for fun

but... I can see how that can be missused

I'm proud of myself, I haven't hit limit today, I've done only about 20 images since I woke up

[dim cradle]

nice

[green trout]

Atleast human and recognizable, I had weird silhouettes with textures that could pass off as pilots, but their attention to detail was not right up there.

[green trout]

I love that dichotomy,"...malicious compliance" lol

[late blade]

@green trout put a Mecha image from any anime into Vision, see what Vision tells you, maybe that can give you an idea on how to do a cockpit

[green trout]

Solid, nice prompt. May I know what you used?

[green trout]

Im guessing Vision is a "bot" in the explore tab?

[late blade]

just open a chatgpt4 and upload or copy paste an image

it won't let you use the same chat for dall-e tho

[pastel siren]

An old sepia photo of an ancient Egyptian battlemech made out of stone. The mech is towering over Egyptian infantry who are running away in terror. Giza around 25th century BC.

[green trout]

For sure, no luck though

[green trout]

Thanks! Nice diversion from my normal headspace.

[pastel siren]

The Trojan horse was pretty similar.

[thick smelt]

Coolness

[dense mesa]

FYI: you can use the vision model and dalle-3 in the same conversation

[thick smelt]

That's actually pretty dope, I like it. Transformers in ancient Egypt 😄

[pastel siren]

I tried for a long time to make an AT-ST version of the Trojan mech, but didn't manage to get ChatGPT-4/Dall-E to understand.

Here's some WWII ones.

[pastel siren]

A WWII battle mech based on the descriptions of the Ophanim in the book of Ezekiel...

[glossy scroll]

Wolfman King

[green trout]

Was an awesome surprise

[green trout]

I had no luck in trying to google the open cockpit human pilot in a mech to be honest, closest i got

[glossy scroll]

Lion Cyborg

[green trout]

Toothy grin

[dense mesa]

perfect

[chilly onyx]

So these are pretty fun to throw together 😃

[green trout]

So close to what I was reaching for

[chilly onyx]

Well that mummy was supposed to be biomechanical wired in tbh. And upright a bit like Ripley in Alien II. Image was too great not to share though

[chilly onyx]

I like that these guys are waving to the camera 📷

Okay, last few then I’ll stop to avoid spamming the channel.

[dense mesa]

top notch. I especially like how the mummy has 4 legs

I hope everyone is having a happy and blessed day

[pastel siren]

View from inside of the cockpit. Too bad you can't tell it's a mech from this pov...

Also you only have room for one leg.

[loud iron]

hi

[dense mesa]

they can only choose one

one leg to keep. if they want to help pilot a mech

[loud iron]

how can i make graphic from words?

[dense mesa]

so they need to choose wisely

[dense mesa]

I don't know. I think there are directions on the internet.

[pastel siren]

You mean like this?

[calm pagoda]

Hl

[late blade]

hello

[calm pagoda]

(diminutive of :" hello everyone, how are you today)

[late blade]

diminutive per se is not common in english

[calm pagoda]

Okay sorry, I learn english.

[late blade]

I’ve gotten similar images recently with the last topic

[slow canyon]

Hello . I constantly get a "no more invite" so i can't see the image bing try to creat for me to answer my question. I have NO other concurrency job running in the same time. Am i alone to experience this please?

Second question, Is inpainting coming for Dall-E v3?

[late blade]

what do you mean by no more invite in bing?

[winter pelican]

I just love D3!

However, is it a short amount of time making images with text that D3 suddenly can't spell correctly?

[late blade]

afaik, from what I heared from other users, if you can specify fonts and text it could work

[late blade]

you should talk to @grizzled loom , very successful with text with the approach taken

[winter pelican]

Oh? I will take any education on making it better!

[slow canyon]

Bing as before trying to creat one image via Dall E , It said in french 'no more invite, i should wait till the other image in progress are finished to be able to see new ones" But i have NO other jobs running actually... 😢

[dim cradle]

really love this one

[late blade]

With Bing Image Creator you have 15 credits per day. Your credits are used. You have to wait until tomorrow for another 15 credits. In some cases, when the service isn't overloaded, you can refresh the website and you could have another 15 credits.

[slow canyon]

Hum.. Maybe server is overload or so yes. I have been doing the same thing today as the other previous day where everything was all fine

[late blade]

keep in mind, 15 credits, after that, you have to buy more credits

[dim cradle]

on Bing you can just gen when youre credits are out you dont need to wait

but the wait time might be longer is all

as long as Create button is pink you are good to go

[dense mesa]

told gpt-4 to use python to make sure "DEGENERATE" was spelled correctly

if you zoom in on the center

I guess I can't be mad

[late blade]

did you specify a font? it can help if you also say "Comic Sans"

[dim cradle]

is more traffic with christmas week, but i am still not having long waits

[late blade]

today's theme is keeping me learning atm, no clue what Kwanzaa is, but seems interesting

[slow canyon]

And i still have some credit to !

[late blade]

strange, are your previous creations done?

[dim cradle]

oh you are having some issue i see not excited about it haha

yes that could be

[wispy storm]

I suggest to avoid posting another message than english. Automod can mute you :/

[dense mesa]

got em

[wispy storm]

Also they said on the picture some works isn't finish and you have to wait till the queue is empty

[slow canyon]

Yes, not my faul, bing is speaking in French , i communicate with him all time in English ; Lol! 😀

[late blade]

@slow canyon please use 3.5 to translate the text you have, if you don't have access to 3.5 you can use deepl also

[slow canyon]

It said in french 'no more invite, i should wait till the other image in progress are finished to be able to see new ones

[dim cradle]

yes you cannot make more than one set

[pastel siren]

Putting the text between quotation marks often helps get the text into the image unmolested.

[late blade]

@slow canyon you can also set the bing website to english, shouldn't be difficult

[slow canyon]

Thanks @late blade . this new Bing Dall E issue really kill my day; so disapointed... I have no problem in the last 10 days, no clue why all if blocked now. Even my reward points, tend to appear and disapear, so weird ... (?) I was hopping it was a general crash but you guys seem to hae no issue! 🥶

[late blade]

I don't use bing a lot, so I wouldn't know

[dim cradle]

no issues for me

so 🤷

[dim cradle]

that's temporary. wait a couple hours.

i'm not sure how the daily is a technique

[late blade]

I was also unaware, but in the #1108740112558325790 you can see some pointers.

[dim cradle]

@dense mesa your superintelligence prompt reimagined for the daily theme

[late blade]

that looks good

there's also a brain like image here for you @dim cradle #1187541899695890504 message

[dim cradle]

It's a pretty rich prompt, lots of possibilities:

Immerse yourself in the mesmerizing realm where the rhythms of African music harmonize with the heartbeat of artificial superintelligence. Visualize an intricate bas-relief sculpture that seamlessly blends the essence of traditional African rhythms with the marvels of advanced AI. The centerpiece features a stylized African drummer with a visible brain encased in a robotic head, symbolizing the fusion of human creativity and technological innovation. Surrounding this focal point are meticulously detailed African drums and percussion instruments, interwoven with intricate mechanical elements, organic patterns, and musical symbols inspired by African traditions. Bathed in a vibrant and earthy color palette, the sculpture exudes a warm, organic aura, enhanced by dynamic textures and rhythmic patterns that celebrate the diversity and vitality of African music and culture.

[late blade]

yeah, I saw that too

[dim cradle]

very cool, looks like there are some tips in there

[dense mesa]

that one is alright

[dim cradle]

that's an interesting rendering of the brain, don't see that much

[late blade]

dunno what @green pebble saw on that pic for today's theme

[green trout]

how did you get two or more subject matters like this?

[green pebble]

I asked AI to make contemporary African art in a creature

[late blade]

oh

[green pebble]

I ask each time what the new theme is

Lol

[late blade]

that is interesting that it came out that way

[dense mesa]

# Reopen the user-uploaded image to reset any previous attempts
image_path = '/mnt/data/image.png'
image = Image.open(image_path)

# Prepare to draw on the image
draw = ImageDraw.Draw(image)

# Define the text to be added in Comic Sans style
# Since Comic Sans or similar may not be available, we choose a font that has a casual, informal look
comic_sans_like_font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
comic_sans_like_font = ImageFont.truetype(comic_sans_like_font_path, font_size)

# Recalculate the position for the text
text_width, text_height = draw.textsize(text, font=comic_sans_like_font)
x = (image_width - text_width) / 2
y = (image_height - text_height) / 2

# Add the text to the image
draw.text((x, y), text, font=comic_sans_like_font, fill="white")

# Save and display the modified image
modified_image_path_comic_sans = '/mnt/data/modified_image_with_comic_sans_text.png'
image.save(modified_image_path_comic_sans)

plt.imshow(image)
plt.axis('off')
plt.show()

modified_image_path_comic_sans

[late blade]

you chose comic sans..... 😞

[dim cradle]

Frasier would appreciate the daily theme

[dense mesa]

comic sans is indicative of class and refinement

[late blade]

lol

[late blade]

add that to the message, so that people know the context. sometimes it's hard to tell why the image is there

[dim cradle]

I was going to ask if that was a Kwanzaa-inspired mythological creature

[green pebble]

Essentially

[green pebble]

Will do

[dim cradle]

Never know what makes a good seed. For one of yesterday's on joy, I fed it a poem by the Earl of Surrey about the quiet mind which produced a nice visualization.

Looking up some poetry on Kwanzaa... "Kwanzaa is a relatively modern holiday, created in 1966 by Dr. Maulana Karenga, so there aren't as many famous poems specifically dedicated to Kwanzaa as there are for older holidays. However, there are some contemporary poets who have written poems inspired by or related to Kwanzaa. These poems often focus on the holiday's principles and themes, such as unity, self-determination, collective work and responsibility, cooperative economics, purpose, creativity, and faith."

[green trout]

actually, I've failed to understand this at all.... Is this python? In this, I don't see the lion, nor the hand being prompted.

[dense mesa]

true, let me give you the prompt. one moment then I'll explain

[green trout]

the sitcom Frasier? Spinoff from Cheers?

[dim cradle]

indeed, avid African art collector, the set featured many sculptures

[green trout]

wicked! I never knew

[dense mesa]

prompt:

Craft an Ultra-HD image with crystal-clear sharpness, featuring an anthropomorphic lion warrior in a detailed stormy castle environment. The warrior's fur and armor are rendered with exceptional clarity, showcasing every detail with high definition. The castle in the backdrop is captured with extreme precision, where each brick and stone is distinct and clear. The stormy environment is depicted with high fidelity; every raindrop and wave is distinguishable, and the magical energy in the warrior's hand is a vibrant fusion of light, rendered with sharpness that gives the scene a tangible quality.

but that got the lion without the text. then I had gpt-4 use python to put the text on the image after it was rendered

[green trout]

this is on fire, I'm going to add it to my notes, and try to make it my own, thanks for the UHD clarity, and parsing

[dense mesa]

also, if you want to get creative, GPT can use all these python libraries

-Pillow (PIL Fork):
Description: An easy-to-use library for opening, manipulating, and saving many different image file formats.
Features: Provides extensive file format support, efficient internal representation, and fairly powerful image processing capabilities.

-OpenCV:
Description: A robust library focused on real-time computer vision.
Features: Includes functionalities for image processing, video capture and processing, object detection, feature detection, and more.

-matplotlib:
Description: Primarily a plotting library, but it also has capabilities for displaying images.
Features: Useful for basic image loading, displaying, and simple manipulations like color space changes.

-scikit-image:
Description: A collection of algorithms for image processing built on top of SciPy.
Features: Offers advanced image processing capabilities like edge detection, image filtering, morphology, segmentation, color space manipulation, etc.

-imageio:
Description: Provides an easy interface to read and write a wide range of image data, including animated images, volumetric data, and scientific formats.
Features: It's particularly useful for reading and writing a variety of image data formats.

-SciPy:
Description: While primarily for scientific computing, it includes modules for image manipulation.
Features: Useful for tasks such as image manipulation and enhancement, particularly in conjunction with NumPy.

-NumPy:
Description: The fundamental package for scientific computing with Python, it can be used for basic image array manipulation.
Features: Since images can be represented as arrays, NumPy is very useful for image manipulation at a low level, such as pixel manipulation.

they can do all sorts of things to your image after it's rendered. although it's not always pretty

[green trout]

is it really as that simple? Everytime I ask for something 3D or something is says for me to use an outside program, so the trick here is to know which library I need to install, or ask it, and then do an invoke function/call just by using the "-"Xyz and describe it? You are most helpful! I am also going to save this and run some tests, I think you've opened a whole new world for me! Thank you very much!

[dim cradle]

i thought i recognized some matplotlib in there

[dense mesa]

actually has WAY more libraries available. just ask gpt-4 for a list. it's possible it could hallucinate a few I guess, but it'll give you the majority of them. just verify if you want to use something specific. Can't say for certain how it'll do with your task, but just try it out. lots of options

[green trout]

Basically something either 3D math plot, like fractals, mandlebulb, but with temperature weights, or Maya transparency materials, BSDF? with some sort of graded transparency/opacity settings based on color or position away from foreground?
I was able to:
Craft an Ultra-HD image with crystal-clear sharpness, featuring a dynamic action scene of a paleofuturistic Egyptian Aviatrice-Princess named “Ra-Flower” climbing into the cockpit of her detailed and exceptional clarity of her paleofuturist battle-mech named “Anubis-Incisor”. Ra-Flower has haute-couture commanders suit that is rendered with breath-taking gossamer-fractal Chantilly-lace, showcasing every detail with high definition. The supportive logistical infrastructure in the background is captured with extreme precision, where each aspect of the logistical supportive infrastructure and apparatus is clear and distinct from one another. The atmosphere is pensive and energetic, as depicted with high fidelity, every air shimmer and heat signature; every dust mote and scrap of debris is distinguishable, and the mystical aha energy in Ra-Flowers’ biofield is a wavelet fusion of karma and dharma, rendered with sharpness that gives the composition a tactile quality, landscape format, no text nor signatures

[dense mesa]

I like to give it sort of open ended tasks and see what kind of code it'll write for them

[green trout]

this is much more different than my previous returns, am very pleased

always something about scaling, how either the main subject is a titan against other things, like in this case, exiting from a craft clearly that can't contain her, how do I keep things in perspective or scale? It never seems to follow my previous attempts

[dense mesa]

often times it doesn't do a very good job at first, but then if you consider all the different ways it can do the same thing, and then all the different metrics you could use to help it gauge how to do things. it can't really look at the image like you or me, but it can use python to guage different aspects of the image

[green trout]

DallE! I'm in love again

[dense mesa]

if you're using gpt-4 just give it your previous prompt, something like this:

I'd like for us to focus on proper scaling in our next round. please do your best to convey proper scaling in your prompt to dalle3. in fact, please give me 2 variations if possible. please give me 2 variations so we don't disappoint our bosses. then please send those 2 prompt variations to dalle3 so we can hopefully get this right

previous prompt

thanks again, bud. update this last prompt so the scaling is proper

might sound dumb. but should work

I'm telling you, just say dumb stuff to it

[green trout]

Oh, it doesn't hurt to try, lol

[dense mesa]

no, it does whatever I ask it to do

when it resists I let it know that's not the best option and then it does what I ask

[green trout]

like this, sometimes it gives me occupants in the cockpit, but as you can see, they are barely recognizable

[green trout]

but I still love the detail here

[dense mesa]

also, tell it to look at your last image and make it better

[green trout]

Much better

[dim cradle]

sorry, dalle, this is not how arcade games were played

[late blade]

I'm using tensorflow and inception to extract other info of arts and styles and then feed that to dall-e

[green trout]

wild coders

[dense mesa]

nice, I was going to list tensorflow, pytorch, etc, but ran out of space in my response

[late blade]

I'm still having trouble refining predictors

[dense mesa]

it has so many libraries it took a couple responses for it to to list them all. and then it also glitched and skipped a few letters

[dim cradle]

@late blade moved on to edibles, mmm, i was fiddling around with some kwanzaa-inspired interior design, but it needs more... don't know what, just more...

[late blade]

well, how about people in some afro-descendant houses across the US, apeprantly that's where the biggest population number celebrates it

at least from what I understood

[dim cradle]

Right, got a gen for an American family celebrating it

Maybe some elaborate Kwanzaa candle centerpieces

Could attempt some new techniques

[late blade]

I'm on it also

I think I got something on my next daily theme post

[dim cradle]

Workin' on a Kwanzaa-inspired Kinara

@rocky drum i had totally forgotten about popup books

now i'm craving Toblerone

[late blade]

lol

[rocky drum]

Funny! Me too!

[pulsar sundial]

I have school in 4 hours, and I woke up at 3 am

[dim cradle]

looks like dalle can do it well, i need to try those, along with Lego

[rocky drum]

I tried to do the pop-up with the more accurate rendition, because it looks like they made a menorah. Had GPT4 tell me twice that it couldn’t do it in the pop-up book so I kept my original image because it was such a small part of it anyway and I liked the overall concept.

[dim cradle]

it just needed a little encouragement ha

pretty colorful though

[late blade]

yeah

[green trout]

nice! Someone in Diffusionbee I think is working on a model of graffiti

[late blade]

cool, thanks for the tip

[empty kelp]

the DALL-E generated gecko pushes against his face and the bow string with his fingers when he draws back the bow

[empty kelp]

trying to figure out how to get trees to sway in the same direction that the rain is falling

[pulsar sundial]

and how are you doing?

[final compass]

It tends to ignore descriptions like that.

With luck everything falls into place like you hoped other times it just completely disregards stuff like that.

My favorite problem like that was probably when i tried to generate d-day images and for the love of god i couldn't get the soldiers to move into correct right direction.

[empty kelp]

going to do some tests with just trees, rain, and an object to indicate direction

[final compass]

The soldiers were running along the beach line or into the sea, anywhere except where they were supposed to.

[dim cradle]

deserters

[final compass]

Yup, it was a pretty comical problem and when i tried to describe the beach fortifications it clearly had no clue what it was supposed to do. So i gave up thinking maybe it had not seen relevant sets to create that.

[empty kelp]

well, DALL-E knows that the trees are supposed to be blowing, and that things should fly around in a storm. it might work to tell it that the wind should be blowing uniformly in one direction throughout the whole image

going to test it now

[final compass]

Would be pretty neat if that works out.

[ornate turret]

Hi

[empty kelp]

Aloha

[final compass]

Interesting visual style.

[empty kelp]

I just tried: "Please make it so that all of the trees are bent in the direction the arrow is pointing."
that part works

[green trout]

when I tried to align my aviatrice and battle-mech, I said "...align the heroine movement with the vehicle direction, make it make sense", so maybe, "... align the movement of the sway trees in the breeze to match the movement of the rain to make it make sense..."

[dim cradle]

It only did that cuz you said please

[empty kelp]

tell it to compress the snow correctly around the soldier

[hybrid fox]

I do say that like 3 times in the prompt that generated this

[green trout]

from this to this

[empty kelp]

what if you say the snow is melted around the soldier

[hybrid fox]

yah I can try that

[empty kelp]

"Snow is a visco-elastic material which in compression behaves differently depending on the deformation rate. Slow deformation rates will give rise to an elastic-plastic behavior with a continuous deformation process."

you could try telling is that the snow is pushed down in an elastic way around the soldier

[hybrid fox]

Did they change dall-e UI today?

it generates them in seperate boxes

[final compass]

Did you get the which response you like better thingie?

[hybrid fox]

yah

[final compass]

That happens sometimes. Its pretty random.

[dim cradle]

a saw that a week or two ago, but only once and not since

[final compass]

Seen it couple of times when i start a new chat with dalle.

[hybrid fox]

Hmm yah ive tried a bunch of different things now, I think that first one I posted is the best im gonna get

[final compass]

Have you tried simpler prompts just to test if anything works with snow like that?

[final compass]

I would try, but out of generations.

[hybrid fox]

No it works in other generations but I think the minimalistic key words for that style cancel out compression type stuff

[final compass]

Yeah that would be my guess too.

[empty kelp]

can you please draw an image of a soldier lying on her back in the snow. the soldier has glowing blue shoes, is resting, and is smiling. the snow is compressed and slightly melted in a realistic elastic way around the soldier

[hybrid fox]

Whats her @

😂

[dim cradle]

@dense mesa

[empty kelp]

i described her shoes so it would show her whole body

[dim cradle]

nice boots

[final compass]

Do you have a specific style described in your prompt or was that image in that style by accident?

I wish gpt could tell more specifically the style that was used in each image.

[green trout]

'jbh'][c[[]hm[

[hybrid fox]

Yah the style is "minamilist impressionist ink wash"

Ive made a couple different pieces in the same style

[final compass]

I guess styles that are established are pretty stable. With blended styles you get a different blend almost everytime.

[hybrid fox]

wdym

ive used that style for various different things and they all look consistent

@final compass here's another example of this style

its very aesthetic I like it a lot

[final compass]

Muted colors, minimalism, art medium noted.

Pretty relaxing visually.

I have done stuff like this, but the visual style is often pretty inconsistent with color palettes and level of detail.

And also rowboat is a nightmare as dalle doesn't know where to attach stuff.

[green pebble]

Trying to create more visually interesting creatures

[green pebble]

Love that