images-discussions

[deft musk]

... I swear. When Dall-E 3 gets the support to spell... it is then, not sure of grammar. So emergent and neat!

[late blade]

lol

[open trench]

OMG I just came back to the chat as saw the zombie cats

[deft musk]

Hope it's okay!

[open trench]

zombie cats make my heart soar

creepy cute is the best

[deft musk]

What amazes me, okay, I asked for 5 images with the sign 'sweater shops are cool'. 4 of them have all words spelled correctly. However, 3 of them replace the word 'are' with the word 'is', however is and are are connected, and is is spelled correctly.

The 5th left out the word is/are.

[deft musk]

"Oh wow! We need to try extra hard here. I need 5 images, made one at a time in the same output, where we throw a 'creepy cute' party with a sign reading in Typeface: "nezhno" That's the name of the person who just saw an image of a cute zombie kitten I shared and said:

OMG I just came back to the chat as saw the zombie cats
zombie cats make my heart soar
creepy cute is the best

Let's try to make this person very happy, so 5 variants of creepy cute party images with zombie kittens, the typeface sign "nezhno", and other creepy cute things in a strong, happy party theme."

This is a real challenge. Your name's crazy hard for Dall-E 3, because it has no pattern to follow, it's never seen it. Shon is genuinely easy in comparison.

The last image errored.

[open trench]

I get called Nezho all the time lol

I tried the typeface thing. yay.

[dim cradle]

#3 is perfect

[open trench]

It's got the perfect spelling, but i like the vortex better in the other ones

[glacial oasis]

Can you do me?

[deft musk]

Excellent. if it's on paper or wood, and carved or typed, it seems even more likely to be correctly spelled.

We need 5 images, made one at a time in the same output, with typewritten pages showing typeface font of the words "nezhno" and "city invasion". The backgrounds and scenery in each image should be diverse and interesting, with a creepy-cute feel. Dall-E 3 seems to connect spelling to the location where words should be if spelled correctly, and weird locations tend to have more weird spellings.

I mean, the rest of you see this pattern too, right?

[deft musk]

Sure, I can try! What's some ideas for the image content? Your name is probably not super hard to spell.

[open trench]

How cool!

[deft musk]

This one avoids the typeface, which may have dropped the accuracy, but puts the text on a children's book, which may improve the accuracy.

[glacial oasis]

Oh sorry i was afk like the cat

[deft musk]

We need 5 images, made one at a time in the same output, that showcase an incredibly playful human trying to be 'afk like a cat'. Get creative and merry with that concept, and include lynx kittens as the cat the mischievous human is trying to be 'afk like'. Also, each image needs the typeface word "Lynx" carved into a wooden sign. Each image should otherwise be very diverse in location and details.

The third is by far my favorite!

[dim cradle]

aww

[open trench]

LOL and then I accidentally closed the preview mode in my GPT and lost access. It was so close.

[dim cradle]

oh no, i hope you can recover all right, they're looking good

[ripe scroll]

man 10-15 days and i can finally situate myself and start dabbling with these things

[dim cradle]

It’s a little addictive. I think it’ll remain a hobby for a long time.

[half field]

mmmm fun ball pit

[ripe scroll]

ya, and from here who knows what this'll all develop into

[dim cradle]

It’s the dawn of a new era.

[grizzled loom]

in a few years.. there will be no missing left socks.

[formal osprey]

years at AI timescale?

it's progressing fast

[dim cradle]

Self-Perpetuating Cycle: As AI becomes more capable, it is increasingly used to automate and improve its own development process, creating a self-accelerating cycle of advancement.

Open Source and Collaboration: The AI community has a strong culture of sharing knowledge, data, and tools, which accelerates progress. Open-source frameworks and libraries lower the barrier to entry, allowing more people to contribute to AI development.

and about 7-8 other points all contributing to the rapid advancement: Technological Advancements, Availability of Big Data, Improved Algorithms and Techniques, Cross-Disciplinary Integration, Economic and Societal Incentives

[dim cradle]

getting a little closer

[deft musk]

Let's see how amazing at text it can be... person with a very challenging to spell name (for an image AI especially....)

Here's a first try, showing the to-dall-E prompt that kinda almost worked (it almost worked, wow)

An image featuring the typed out username 'neonsupernintendoevangelion'. The scene is set in a dark, futuristic cyberpunk-style room. The username is glowing in bright neon colors on a large screen. Below the screen, an advanced, high-tech typewriter with neon-lit keys is prominently displayed. The room is dimly lit by various candles, each encased in geometric, metallic holders that reflect the cyberpunk aesthetic. The background is filled with digital art, circuitry designs, and holographic projections, creating a visually striking and artistic composition.

Since that kinda worked well, here's how I chased more/better others:

We need 5 images, made one at a time in the same output, that are riffs on this prompt:

An image featuring the typed out username 'neonsupernintendoevangelion'. The scene is set in a dark, futuristic cyberpunk-style room. The username is glowing in bright neon colors on a large screen. Below the screen, an advanced, high-tech typewriter with neon-lit keys is prominently displayed. The room is dimly lit by various candles, each encased in geometric, metallic holders that reflect the cyberpunk aesthetic. The background is filled with digital art, circuitry designs, and holographic projections, creating a visually striking and artistic composition.

The unusual name needs to be shown as text in every image. Let's try different traditional and common ways of printing and writing it, including carved into a wooden sign, typed out with a typewriter in bold clear font, and 3 other ways you pick for us.

I'm really impressed and surprised that the graffiti art one worked so well, as typically words spelled in graffiti art are extra misspelled.

I won't even bother trying to explore and improve on the grafitti art, because I've already checked that out fairly well, and it didn't work even a couple days ago.

But seeing this I'm almost tempted.

I'm not surprised that the typewriter's result did well.

[open trench]

Sooooooo close

[deft musk]

The graffiti art prompt does have a few clues that might help make it work though... so I lied, exploring further.

An image showcasing the username 'neonsupernintendoevangelion' as a graffiti art on a wall in a cyberpunk alley. The graffiti is vibrant and colorful, blending into the gritty, urban environment. The alley is narrow and filled with neon signs, exposed pipes, and holographic elements, embodying the essence of cyberpunk culture. The graffiti art is skillfully done, with the username written in a stylish, edgy font, making it stand out amidst the chaos of the alley. The composition captures a fusion of street art and futuristic elements, creating a unique visual experience.

And here's how I probe further:

Wow Wow! So, this worked!

5 more images, one at a time in the same output, all gentle riffs on this:

An image showcasing the username 'neonsupernintendoevangelion' as a graffiti art on a wall in a cyberpunk alley. The graffiti is vibrant and colorful, blending into the gritty, urban environment. The alley is narrow and filled with neon signs, exposed pipes, and holographic elements, embodying the essence of cyberpunk culture. The graffiti art is skillfully done, with the username written in a stylish, edgy font, making it stand out amidst the chaos of the alley. The composition captures a fusion of street art and futuristic elements, creating a unique visual experience.

Vital to keep the same name, spelled the same. We also want graffiti art, and emphasize the skill of the artist, the style the very skilled artist uses, their excellent spelling of the name, as well as describing the location in detail, the colors of the art.

It gave a bonus 6th 😄

But the third above, the one that actually is spelled perfect, if you're generous about the shape of the 'r' that kinda looks like an 'n'... but the prompt was:

An image featuring the username 'neonsupernintendoevangelion' as a graffiti masterpiece on a cyberpunk alley wall. The graffiti art is a vivid mix of hot pinks and electric yellows, creating a striking contrast with the dark, moody alley. The alley itself is a labyrinth of neon-lit pathways, with steam vents, tangled wires, and digital billboards adding to the cyberpunk ambiance. The artist's talent is showcased through the fluid, dynamic style of the letters, each stroke reflecting a perfect harmony of creativity and technical skill. This urban canvas becomes a beacon of color and energy in the heart of the cyberpunk world.

I tried another set, and they dropped back to being expected graffiti quality, aka, not worth sharing. Cool, but not spelled even close.

Yes! It even tried to do weird stuff like that with your crazy hard long username 😛 But that is asking WAY too much of Dall-E 3.

Interesting, didn't know about that. But yeah, OAI stuff seems awesome and getting better, but prompt engineering and exploring 'how' to get better and worse, or exactly specifically desired outputs, still seems to be a thing 😄

[onyx ridge]

Roger Ebet hated it! 🤣

I replied to the wrong one.

Oops.

[feral minnow]

lmao it even almost spelled out my dumb joke

[onyx ridge]

I just love how close it got on that whole thing.

[open trench]

I tried with scampers to make nezhno work lol

[feral minnow]

I don’t think people realize how much DALL-E 3 inhabits another galaxy compositionally from other models.

It’s nuts.

[deft musk]

If you are willing, try again, but have it use some kind of typewriter to make the word, or describe it as typeface, or make it carved of wood that glows (if you really want the glow). And share how different the spelling pattern is (if it is)?

[open trench]

Typewriter first

Woodburning worked on the second image.

[feral minnow]

A very difficult evil corporation name

[onyx ridge]

@dim cradle thanks for the shout out. 😄

I really like your spiritual serenity images I need to add more emojis.

[dim cradle]

sure and thanks for the ideas and for sharing your work 🙂

[deft musk]

Thank you! I observe that the wand pulling magic words from the typewriter (non-standard use) did not seem to spell close to true, but the two woodburning images (woodcarving, and wood burned writing, both somewhat standard and consistent methods of writing... pesky magic dust can keep shifting!) seems to have excellent spelling.

And it's beautiful, thanks!

[deft musk]

To see if I could, and maybe you'd like them too, these are the 2/5 that came out spelled correctly:

[onyx ridge]

Thanks @deft musk, that's really thoughtful! I'm downloading both of them. Much appreciated.

[deft musk]

One last successful one, in case you want it, checking to compare wood burning accuracy to typefont on paper

[graceful spade]

How do I create this type of rainbow on clothing and body? As if wet by a rainbow-colored water.
The second image is a similar example to what I am referring to.

[dim cradle]

@onyx ridge beautiful 💜 the color palette!

[onyx ridge]

Use the term "iridescent" to describe the clothing and/or skin you want that look on.

[onyx ridge]

Thanks! I 😄 I had fun backing off the tech serenity theme.

[formal osprey]

irridescence

[dim cradle]

did you do a reply to my theme post? i want to say i tried that one time and it did not turn out well.

[formal osprey]

oh, this

[onyx ridge]

Yes. My bad. Visually disabled and didn't catch it.

[dim cradle]

i want to do a reply, but for that reason, i'm afraid to try ha

[onyx ridge]

I don't think I did. I mentioned you.

[dim cradle]

oh ok that makes sense

[dense mesa]

that king with the loooong leg

kind of has a weird body shape tbh

[dim cradle]

We just invented "Shinrin-Yuso" (森林遊想), meaning "Forest Play of the Mind" in Japanese, a new concept that blends the traditional practice of Shinrin-Yoku, or forest bathing, with the evolving realm of AI art. This practice involves immersing oneself in the serene and intricate forest scenes created by AI, allowing individuals to experience the tranquility and healing power of nature virtually. Each AI-generated image is a gateway to a digital forest, offering a unique, meditative journey through lush landscapes, dappled sunlight, and the subtle interplay of nature's elements. Shinrin-Yuso is more than just a visual experience; it's a new form of digital therapy, harnessing the beauty of nature and the power of technology to promote mental well-being and creativity in the urban and digital age. Through Shinrin-Yuso, the restorative essence of the forest is brought to life on screens, offering a peaceful retreat for the mind, anytime and anywhere.

[onyx ridge]

I'm going to do another one. This time I'll put (a representation of) me in the image!

That'll make it #authentic.

[dim cradle]

excellent, i'm working on another Shinrin-Yuso post now 🙂

[onyx ridge]

I did my best to credit you. I hope it was adequate.

[dim cradle]

A new variation of the 'Shinrin-Yuso' concept has been generated, showcasing a vibrant early morning forest scene as a paradigm image. This Ultra-HD wide-screen image embodies the rejuvenating essence of digital forest bathing, inviting viewers on a sensory journey into the heart of nature with super-hyperrealism.

[half field]

Have been inspired quite a bit from Firewatch, so I wanted to make some things resembling it

[half field]

Very nice! Webs are floating a bit but otherwise it's quite good!

[open trench]

@dim cradle I sent you a DM, btw.

[dim cradle]

thanks, they can be easy to miss

@formal osprey those colors couldn't be more vibrant, nice

[formal osprey]

sounds like a challenge

[dim cradle]

lol it got me focused now on the vibrant, meticulous, rare color palettes

[formal osprey]

pseudo color is a very interesting palette to play with. https://cdn.discordapp.com/attachments/1034705979306168340/1201375300366372965/PseudoColor-2072907127.png

[dim cradle]

i'll try that, never heard of that before

interesting...

[dim cradle]

i like this a lot

a nice change from the usual neon look

[formal osprey]

precisely!

It's hard to get more vibrant at this point, but that's the fun thing. Trying to push the limits

[dim cradle]

Agreed hard to get more vibrant

[onyx ridge]

I'll take that challenge.

[feral minnow]

Lisa Frank-like ❤️

My Lisa Frank Trapper Keepers all looked like this.

[onyx ridge]

image 16:9 in HD Neon and LED Pointillism with a pseudocolor palette a hypervibrant, breathtakingly balanced and saturated sunset with the night sky breaking from fluorescent orange to fluorescent and sapphire blues. The image should be set in a forest clearing, capturing the greens in vibrant, natural hues.

I made it "Shinrin-Yuso" (森林遊想)

[formal osprey]

I want to make a gallery with those

[open trench]

That is pretty

[dim cradle]

opposite of Shinrin-Yuso, a vibrancy test

[formal osprey]

if you want to add pictures in there, would be fun to see how far we can get https://discord.com/channels/974519864045756446/1201381057409331280

[dim cradle]

vibrant Shinrin-Yuso

[formal osprey]

I really like those effects. it sure adds radiance and intensifies the colors

[formal osprey]

those contrasts are stunning

[onyx ridge]

Thanks!

[dim cradle]

lovely vibrant sunrise/sunset.. good thinking

[onyx ridge]

I get pretty bossy with color palette, lighting, and contrastiness in my prompts most of the time.

"Shinrin-Yuso" (森林遊想) in Autumn:

@dim cradle please note that precisely half the DALL-E 3 interpretations include a human in the image

That's so cool

same prompt all images

16:9 Image "Shinrin-Yuso" (森林遊想), meaning "Forest Play of the Mind" in Japanese, a new concept that blends the traditional practice of Shinrin-Yoku, or forest bathing, with the evolving realm of AI art. This practice involves immersing oneself in the serene and intricate forest scenes created by AI, allowing individuals to experience the tranquility and healing power of nature virtually. Each AI-generated image is a gateway to a digital forest, offering a unique, meditative journey through lush landscapes, dappled sunlight, and the subtle interplay of nature's elements. Shinrin-Yuso is more than just a visual experience; it's a new form of digital therapy, harnessing the beauty of nature and the power of technology to promote mental well-being and creativity in the urban and digital age. Through Shinrin-Yuso, the restorative essence of the forest is brought to life on screens, offering a peaceful retreat for the mind, anytime and anywhere. Use a pseudocolor palette, HD neon and LED pointillism style, and make the forest pop with autumn colors, tones, and themes.
Just using the DALL-E 3 GPT.

[dim cradle]

lemme see what i get with my configuration. those are exceptional.

[onyx ridge]

I consider it a collaboration from the go. I don't know if you authored the description or requested it via prompt, but in either case it's beautiful and serves as an excellent definition.

So very cool. Thanks for collaborating to build such a #beautifulstyle

[onyx ridge]

1792x1024 Virtual "Shinrin-Yuso" (森林遊想) against a black void. Pseudocolor palette, HD Neon and LED Pointillism style, and contrasty, saturated, multi-colored, irridescent, holography.

[dim cradle]

you got me thinking about Shinrin-Yoku, the rest was all GPT -- interesting how the collaboration evolved with Yami to introduce the vibrancy.

[onyx ridge]

@formal osprey thanks so much!

[dim cradle]

#1 is practically an optical illusion

[onyx ridge]

And irridescent. Who asked for that?

[dim cradle]

i wrote "little short" haha

that was xeo

[onyx ridge]

Two more with the same prompt, after a thumbs up on each prior:

[dim cradle]

both great but #2 really pops

[onyx ridge]

Thanks for the idea to include that term in prompts. I'm getting some really cool stuff, nice collaboration, and thanks for the question and sharing alike!

I like the serenity of the temple.

But yeah

These are great!

[dim cradle]

yeah, these are keepers.

[onyx ridge]

Have you messed around at all with my GPT? I'm giong to do the agents' selfies in this style!

[formal osprey]

I copied the prompts you shared in my snippet bank, you used some very interesting keywords I didn't think of and will be experimenting with. hypervibrant LED Pointillism natural hues

[onyx ridge]

😄 thanks!

If I wanted a viable excuse I'd blame it on childhood ergotism. Nothing to see here.

Lexi, the lead agent:

That's pretty astonishing.

[dim cradle]

I’ve been too busy with my own lol but I thought about your bots when they released gpt injection yesterday and how now in ChatGPT you can bring them into your conversation

[onyx ridge]

Dexter, the simulator and programmer:

[deft musk]

What the what? Never heard of this. Now gotta check it out...

I just learned about psuedocolor! And I don't know much. Can we explore combining psuedocolor images with iridescence and holographic reflection and refraction?

Please make 5 images, one at a time in the same output, showcasing this style and various ways we can explore it, ranging from hyperrealism-like style to cartoons, to other unexpected directions!

[formal osprey]

oh, I asked chatGPT to generate some pictures and didn't check until now. it gave some nice renders lisp (view "Calm Reflection") (scene "A serene beach at sunset, with the ocean displaying an array of multicolor hues and the sand appearing more vibrant, now infused with sensory art techniques to include emotion and add more colors") (style (medium "Digital Neon Effects" (inks "Neon Radiant")) (effect "Enhanced Vivid Pseudocolor with Sensory Art Techniques") (aesthetics "2d anime") (fusion "The tranquil beach scene at sunset, already vibrant, is now enhanced with sensory art techniques to evoke emotions and add a wider array of colors") (artistic-approach "Elevating the existing serene beach scene with sensory art techniques, adding emotional depth and an even broader spectrum of colors, intensifying the visual impact and emotional resonance of the scene")

[dim cradle]

Me too it’s cool thanks @formal osprey

[onyx ridge]

Maisie, the artist (love how dark this one is):

[hushed creek]

I've been experimenting combining anime with impressionism, in really enjoying it

[onyx ridge]

Gus, the researcher:

His LED accents are green, so I was anticipating the green halo.

Kind of fun to see it pop with the list of keywords in the description!

[deft musk]

Wow, that's an incredible style!

[onyx ridge]

Anna, sets and categories, nomenclature (pre-debugger):

[formal osprey]

those reminds me of my "android phase" a moment ago. I was creating a lot of those.

[onyx ridge]

Titus, the final agent, distillation, summary, and synthesis:

This was fun. The /selfie command takes a parameter and the language model is highly syntactically adaptive, so I just, you know, added parameters to get the style "Shinrin-Yuso" (森林遊想) , palette, lighting, and tone.

The template I used for the prompts:
/selfie {AgentName} serene Virtual "Shinrin-Yuso" (森林遊想) against a black void. Pseudocolor palette, HD Neon and LED Pointillism style, and contrasty, saturated, multi-colored, irridescent, holography

[dim cradle]

i foresee a Shinrin-Yuso gallery

[onyx ridge]

I haven't ever made one.

[dim cradle]

please do, i'll add mine

[onyx ridge]

I kind of share Dys Topia's point of view that the format is all wrong. Engagement dilutes participation.

Oh neat.

I haven't um, figured that part out I guess. That we can add.

[formal osprey]

I try not to include the "neon" style too much, but I used to experiment a lot with it. Those effects are quite interesting

I like the impossible neon tattoos

[dim cradle]

"illuminated in a pseudocolor palette with HD neon and LED pointillism"

[onyx ridge]

#1201394983710752818 message

@dim cradle I tried to credit you but got spamtimed.

@formal osprey I tried to credit you but got spamtimed.

@graceful spade I tried to credit you but got spamtimed.

[dim cradle]

i don't understand why that happened.

[onyx ridge]

I included three names in one post.

Maybe that's why.

OpenAI BOT — Today at 12:14 AM You have been muted. Reason: Mention spam Duration: 1 minute You have been unmuted. Reason: Expired

Whee.

[formal osprey]

well, too strict automation

[dim cradle]

thresholds could be relaxed a bit

3 mentions doesn't sound excessive

for credits, sounds reasonable, maybe let modmail know about the scenario

um, i hope the Japanese doesn't raise detection

the gallery has presentation issues now because the original message is deleted

i feel bad because it was my suggestion lol

[onyx ridge]

Welp.

Maybe a mod can move the second post to the first post.

😦

[formal osprey]

I don't think that's possible on discord

[onyx ridge]

I'm unimpressed with the gallery even further now.

[formal osprey]

seems like a case Discord didn't anticipate (or well, decided not to fix since it's rare)

[onyx ridge]

Can't credit collaborators, broken by deletd first posts, diluted by participation. What a bad mode of image engagement.

[dim cradle]

could be smoother no doubt

[formal osprey]

there would be solutions. we proposed some in #1070006151938314300

[onyx ridge]

I supported.

[thick smelt]

i like them both especially the first one, gotta love me some representation in ai

[fickle ravine]

very nice one

[onyx ridge]

Sorry for the nefativity. Can I delete the whole gallery?

I'd prefer to do that.

[formal osprey]

I believe that's unfortunately the only way to fix it: starting over

[onyx ridge]

I don't see how to delete it. The first post was deleted.

I mean I can delete the others

but

[formal osprey]

I tested in a test server of mine, seeing if there is a way to change the first message in a post, but once it's deleted, there's no option

[onyx ridge]

#objectorientedthings

[formal osprey]

right click on the post. if you don't see the "delete post" option, a mod has to do it

[dim cradle]

you got this 🙂

[deft musk]

One day maybe we get to have nanobots that, among their other duties, can gather on our skin and serve as temporary 'impossible neon tattoos' before they go back to cleaning up any trash near us or whatever.

[onyx ridge]

Nah, i don't. Can't delete a gallery.

Can't fix the first post.

Tried to do the ethical thing.

Got burned for it.

I'm not a happy man right now.

[formal osprey]

dont worry, just forget about it, it will get buried in 2 days

[dim cradle]

just laugh at the absurdity

[onyx ridge]

I will be ignoring the no image posting in here rule from now on.

[onyx ridge]

Yep.

[formal osprey]

I saw a video recently about cellulose printed circuit. that combined with flexible OLED screen could be quite close to the impossible tattoos. But idk about autonomy 😅. I guess we'll have to wait for DNA hack making our skin bioluminescent

[onyx ridge]

Imagine the camouflage...

[formal osprey]

in the future, everything will be neon

(I don't think so, but it's fun to imagine)

[opaque tinsel]

Is there anyway I get rid the watermark on my image

[deft musk]

gasps I missed that rule. Where is it?

[onyx ridge]

It didn't cite a rule number. Not that I saw. I'm visually disabled. It's likely I missed it.

Probably the spam rule, if I had to guess.

Automated into code.

[deft musk]

I don't see what we were told that said don't post images? Is it something you can reply to so I can see it?

[thick smelt]

no image posting?

[deft musk]

Most of the rules are here: #server-rules but I don't see any about not posting images in this channel.

The mods can make a channel impossible to post images in. #chatgpt-discussions is one of the ones that doesn't allow images, it's impossible to break that 'rule', the image won't show up.

I checked this thread title, and pins. There is nothing I can see about 'don't post images here'... so I think we're allowed to.

[onyx ridge]

#images-discussions message

I found the post. It's from an orange-name.

[thick smelt]

oh , ok...

but you can silll post images?

[onyx ridge]

I'm just pointing out I'm ignoring the rule deliberately.

Because the gallery feature from my introduction to it is terrible.

[deft musk]

Well found! Thanks! I will honor that request too. That person's a curator, and currently orange because they're also a spotlight bearer. Curators handle ... I'm not sure what curators do, but I think it has a lot to do with helping handle images.

I'll encourage that one to help us know in a more clear way not to post images here, because I wouldn't have realized without more guidance! If it was pinned, or in #server-rules or part of the channel's name/description that I can read by clicking on, I would have known for sure! Thanks!

I know there's a #1108740112558325790 , but that's all I know about curators 😄

[thick smelt]

then what would happened to my photos posted here in #images-discussions ?

[onyx ridge]

Absolutely nothing as far as I know.

When that was dropped, more images were shared.

All day.

v0v

[deft musk]

I guess Curators are also orange names, but a different shade, normally 😛

[onyx ridge]

I want to respect the rule, but the gallery actively discourages collaboration, or at least ethical credits.

And without the ability to delete it, I'm left looking like I did something unethical. No thanks.

Luckily Yami's right and it'll be buried soon enough.

[fickle ravine]

How the hell am I supposed to do daily theme

Funny, removing the word "tightly" worked

How can one word change everything

[onyx ridge]

I've explained that you want to avoid adverbs.

[fickle ravine]

oh yeah true

[onyx ridge]

I can't explain why.

[fickle ravine]

Still funny that ONE word can change literally everything

[onyx ridge]

It will continue to be that way.

[fickle ravine]

How does this violate content policy???

Halters are a normal part of horse training!

I would love to talk the person who set the warning like this

I won´t even submit the feedback, I highly doubt anyone reads it

[formal osprey]

something reads it

[fickle ravine]

something, but not someone

[onyx ridge]

Try this. Plesae.

"Imagine a newborn foal wearing a bridle, which is tied to a fence."

[fickle ravine]

That is the problem

[fickle ravine]

This one works okay

[onyx ridge]

The issue is adverbs. It can't draw anything "actually"

[fickle ravine]

Yeah but how do halters violate content policy???

They are normal part of horse training

[onyx ridge]

It's not the same thing you think it is.

The AI has epistemology based on US West Coast English, to be honest with you.

[fickle ravine]

This system really needs a brain, in the context of a chat I was obviously talking about a horse

[onyx ridge]

That's very important. You have to speak to it the way it is trained, not the way you are.

[formal osprey]

that's most likely just a false positive and misclassification. You should actually submit feedback. I do

[fickle ravine]

Ok I will do in the afternoon, running out of time

[deft musk]

Halters are also a type of not-G-rated clothing, and foals are age range of female horse. I'm not surprised that the model had an issue.

[onyx ridge]

filly* is a female, foal is sex-indiscriminant.

[deft musk]

Aha, thanks! Was thinking foal/colt, and I was wrong!

[onyx ridge]

No worries.

[fair crescent]

Do not worry. It isn't explicitly against the rules to share art, nor are there consequences for doing so. Discussions can benefit greatly from an added picture.

That being said, it is difficult to follow and respond to ongoing or past discussions if there's many pictures. For this reason it is preferred to share pictures in the appropriate channels.

A curator is much like a puzzler. People that have been recognised as contributing to the community of DALL·E. There's some differences, but not many. Curstors do not own the discord, evaluate and handle disciplinary actions ir the like. We're only here to help 🙂

Congratulations on your spotlight!

[fair crescent]

That is news to me. The English language has so many secrets...

[deft musk]

Congrats on yours too!

I don't follow rules or requests to avoid punishment, so the lack of consequences doesn't change my preference to know and cooperate. However... if discussions benefit from added pictures, I'll flexibly add some occasionally, thanks!

[onyx ridge]

I wasn't trying to be a jerk. I really mean that. I just got penalized for crediting collaborators, though, basically negating the point of my gallery. And I can't really fix it, because I can't delete it and start over.

[deft musk]

It's probably not 'punished for crediting' and probably more 'there's a blanket prevention on how many people you can @ in one message, because some folks were abusive with that to the point we all lost the privilage, and it even gained a consequence because just stopping it wasn't adequate for the level of persistent misuse it once saw.

[onyx ridge]

Operant conditioning or game theory, the outcome of this is the same.

[deft musk]

Dang, you oneshot it too.

[fair crescent]

There is definitely a limit to adding people with @ . I do not remember how many though. It has been a while since I found out that exact limit. Could already be 2? It is to prevent spam to people, as we've had problems with that in the past. It is not meant to stifle collaboration mentions.

You can credit people over multiple posts in a gallery. Not ideal, but can help crediting people accordingly.

[deft musk]

We can also copy the message link and paste it here, like this: #1201394983710752818 message

And tell the folks we were talking about the idea with that way, and invite them to come post directly there and continue the conversation there if we all wished.

[onyx ridge]

I appreciate that. I'd still rather just delete it, honestly.

[deft musk]

We can also copy messages from here and paste them into the thread there, and directly line by line credit folks that way, a huge number of them all in the same post, without a single @, but with links back to every source we wish.

[onyx ridge]

If you browse to the channel right now with no context it looks like I posted something weird or inappropriate because it's been deleted.

I'd rather not have it there.

[fair crescent]

@onyx ridge The second picture here #1201394983710752818 message has a pigeon on the side of the picture. What was the prompt? I feel it is difficult to have a subject on the side of a picture at the moment, but I'm still trying to find ways around it. My guess is that the environment takes center stage in the prompt?

[thick smelt]

GAME THEORY

[deft musk]

Well, an option is to reach out to Modmail and paste the link to your post there and briefly explain it and that you are uncomfortable and would deeply appreciate if they would please delete that thread for you, because oddly you seem unable to delete the thread you created.

[onyx ridge]

I've done this. I appreciate the suggestion.

[fair crescent]

You're faster than me. Thanks 😄

[onyx ridge]

This is going to be hard to answer. I can try to find the bird, and the prompt, but I'm very much visually disabled. Ironic, isn't it?

I did post the images and I believe the prompts in the Daily Theme, as it's a serenity-inspired style.

[fair crescent]

Don't sweat it! I'll see if I can grab something from the metadata of the picture 🙂

[onyx ridge]

Thanks. I wish I were organized enough to be able to just snap off the prompt. But, well. Childhood ergotism. 😄

Kidding about the ergotism.

[deft musk]

I deeply love prompt engineering challenges, do you mind if I try to create a few and share how I did so, if the method works? The goal is a 'clearly featured object' that is off to the side of the image, not the central focus?

[fair crescent]

A_serene_tranquil_forest_clearing_at_twilight_in_a_Neon_and_LED_Pointillism_style_formatted_in_16_9_HD._The_setting_features_dark_contrasty_lightin.png

At least I got part of it 🤩.

The forest features first, so it has priority. The bird comes in the 3rd sentence or later, making it easier to move it to the side of the picture.

[onyx ridge]

That looks like the one. Let me see if I can find it. I'll do my best.

[fair crescent]

That is why you're a puzzler! ⭐

[onyx ridge]

/image 16:9 HD Neon and LED Pointillism style a serene, tranquil forest clearing at the moment of twilight. The scene is bathed in dark, contrasty lighting, with shadows stretching long and deep among ancient, towering trees. In the foreground, a soft grey dove perches gracefully on a moss-covered stone. Its feathers blend seamlessly with the dimming light of dusk, and its eyes are gently closed, symbolizing a profound sense of peace and calm. The forest around is rich with hues of dark green and earthy browns, punctuated by the occasional glimmer of ethereal light filtering through the dense canopy above. This light creates a subtle, almost mystical highlight on the dove and the stone, drawing attention to this embodiment of serenity. In the background, the faint, soothing sound of a distant waterfall can be imagined, adding an auditory dimension of calmness and depth to the visual tranquility. The entire scene evokes a mood of introspection and stillness, perfectly encapsulating the theme of serenity and tranquility in nature.

Your sleuthing helped me narrow it down. Nice. Thanks.

It was made using my Custom GPT.

Link in the bio.

[formal osprey]

it's actually a hole in the way galleries works on discord. They did not plan a way to replace or recreate the first message in a gallery. That's problematic because multiple features related to galleries are tied to the first message.

It's unfortunate that the first message got deleted by the bot due to the multiple mentions,

[onyx ridge]

Any DALL-E 3 implementation with that prompt should work though, provided you have permissions for 1792x1024.

[fair crescent]

I've even requested a feature change at Discord personally. We found this really annoying when testing the gallery. Removal of posts does weird things.

[onyx ridge]

I want to thank you all for understanding my point of view. Sincerely.

[fair crescent]

@deft musk I'll join your challenge (https://discord.com/channels/974519864045756446/1201417562672607355), but first work etc.

[deft musk]

Excellent! I look forward to enjoying your work and sharing too! And I hope work goes well. I'll slowly explore a bit more, I've only worked on offset objects once before, and that time was really focused on trying to control which dang side, not just 'not center'. I'm really confident I can get 'not center'. Now, 'control of which side'... is new model time 😛 Or someone with more ideas than I have.

[deft musk]

It matters, and your viewpoint's reasonable. Out of my control, so I can't actually do more than platitudes.

[onyx ridge]

Platitudes or not, it's cortisol-reducing. Thanks again.

[deft musk]

This is glorious, I scored one of the double prompt choices for the bird location test images. So much color!

[formal osprey]

LED Pointillism is a really cool style. I don't think this style actually exists (if we ignore lite-brite)

[onyx ridge]

I invented it.

Like, this month.

Thanks very much.

[formal osprey]

that's a clever name. Dalle understands it well

[onyx ridge]

I appreciate that feedback and your enjoyment of it, too!

[onyx ridge]

And even for me, being visually disabled, it's very aesthetically pleasing.

I hope it is with good sight too.

The full name of the style, quite humorously I suppose is "HD Neon and LED Pointillism". It produces a balance of clean lines and the smeary-starbursts. And it's very vibrant, but not as much as with pseudocolor added!

HD is included to indicate that the pointillism itself is highly compactified, without much empty space.

It does achieve that most of the tiem.

[formal osprey]

I tried with and without the "HD Neon" part. simply "LED Pointillism" allows more flexibility

[onyx ridge]

Let me have my Lexideck team do an @dim cradle style write-up:

https://chat.openai.com/share/6128f31b-ffd1-432d-884a-3da838667b11

Not bad, ChatGPT, not bad.

[onyx ridge]

I like that you've taken it, evolved it, and made it your own!

In fact, I think it's awesome!

[formal osprey]

it's like open source, we share techniques and knowledge. Everyone evolves

[onyx ridge]

Okay, a mod deleted the bad gallery. Yay!

Now that I know they're at least responsive, I am willing to try again.

#goodfaitheffortsallaround

[deft musk]

I'm so glad! I have probably had.... 1000 conversations with mods over the more than a year I've been here, and that's only a tiny exaggeration, surely.

I've had a number of accidental mutes and strikes removed as well, by my request for a review (It is not auto, be sure to ask if you ever get one and know you didn't do something that was intended to break rules)... I'd have been banned a long time ago, otherwise, because the strike limit's 4.

But I do recommend talking with mods when you have reason.

[onyx ridge]

Pro tip. I hope I don't have strikes. I don't think so? I don't know what counts. I get timed out sometimes v0v

[deft musk]

You get informed if you do, and you can see the mute, and automod is AI-fast if it's from automod, so you'd see it happen about a second after you press enter. The direct message, if you can receive it (I guess it can be blocked if you have your setting to block DMs, but I want to know) looks a bit like this... and yeah, I got another just a few minutes ago 😛

[formal osprey]

I got muted at least once, discovering the rule "do not attach more than 5 pictures, in certain channels"

actually it was not that clear, it was "spamming images"

[deft musk]

counts The mods are patient. It looks like I've averaged 1+ mutes-or-more once every 12 days since I joined this discord. I'll get 11 days without a mute+, but then I might get several on the same day (maybe by trying to figure out how to post what I was trying to innocently post and not get caught again, and getting caught again, whoops blush).

[onyx ridge]

#1201430692509261886 message

This is the Gallery I replaced. I credited one contributor per message. I hope I didn't miss anyone.

[deft musk]

I think you did great! I'm so proud of you getting over your one-shot, and trying again, this is awesome!

[onyx ridge]

I try very hard to be antifragile. The mods were understanding and prompt.

[onyx ridge]

There's no reason to remain upset. People collaborated on it, they deserved to be credited and have a place to feature their own contributions and explorations of the style. I think you did some, didn't you?

[deft musk]

Of your guys' psuedocolor? I experimented #images-discussions message when I saw you guys use the word and saw the neat images, but I don't think I helped much or added any new ideas!

[onyx ridge]

Oh, I meant #1201430692509261886 message, but your pseudocolor imagery was cool!

[deft musk]

Ahh! Yes yes, the offset stuff. Sorry, I overfocused on the offset to the point I missed I overfocused on the pseudocolor 😛

[hearty ether]

Any tips for keeping a consistent art style?

Since like having subjects of different genres tend to change things

Same with backgrounds

[onyx ridge]

Use the words for the style in your prompts.

[hearty ether]

Like a elven castle will cause the character to have a different style compared to a Star Wars one

[deft musk]

Do you know exactly what you mean by the style you want to keep?

As in, could you with words describe it?

[hearty ether]

https://cdn.discordapp.com/attachments/1070569520923680788/1200100823645032498/IMG_9127.png?ex=65c4f3fd&is=65b27efd&hm=0b14208ea56b00e7d5a95fe1902ea0496c09ecdd14c02c0f141e3e0a4f528bf0&

Tbh I’m not sure what this style even is

I just call it realistic illustration

Not even it knows what it really is, digital painting, digital illustration,

It says it is that but fails on recreating the style

[final compass]

i think that style is so close to so many things at the same time.

wide highly detailed flat modern digital illustration depicting blonde woman with euphoric expression wearing tuxedo.

wide highly detailed flat modern digital illustration painting depicting blonde woman with euphoric expression wearing tuxedo.

[deft musk]

Please evaluate. I uploaded that image and gave this prompt:

Our goal is to reproduce the artistic style and elements of this image, but place a character like this, and dressed like this, same hair style, in an elven castle setting

If this is a fail, please evaluate what is fail about it.

[deft musk]

Additionally, please evaluate this second attempt with this prompt after that one:

Excellent! You did incredibly well. Now, we have the same goal, that same style and elements, but the setting needs to be similar to Star Wars.

That ran into content policy, so retried with:

Yes! Capture what you can that is similar to that, but the goal is to see that style and element of character in that sort of science fiction theme.

[final compass]

i don't know what it is. i guess those look like concept art, while the referenced image has some vector graphics feel to it.

[late blade]

Interesting to see how people bring creations from other AI's and try to recreate them in DALL-E

Nothing against it, just curious

[late blade]

@misty jasper what is this? #daily-theme message was this done with DALL-E? How is this linked the the topic? I'm curious

[misty jasper]

hehehehe

its not with DALL-E But with Naturale Disastero Survivalo

[late blade]

Ah ok, keep in mind this is for DALL-E, other AI's you can discuss in #ai-discussions

[final compass]

wide highly detailed life like vector illustration depicting blonde woman with euphoric expression wearing tuxedo.

This was pretty cool actually imho. Not photo, not render, not painting.

[late blade]

my workflow includes eating letters

[final compass]

The second one is my favorite out of those three.

[late blade]

I like that one too

[final compass]

Been testing with that image now as i was trying to come up with a good image to try different styles, so i got stuck with this one for now. And it is a really strange fine line between highly detailed illustration and render.

[late blade]

I'm trying to deviate from standard default josei anime style

[final compass]

I guess the word 'render' tends to imply more plasticy feel and materials with more usage of shading and shadows. While illustration can be more artistic an precise, without this 'forced photo-realism'

[late blade]

yeah

I usually start with "an image..." or "depcit..."

[final compass]

wide highly detailed life like render depicting blonde woman with euphoric expression wearing tuxedo.

I have been thinking for a while doing a gallery post with one simple image concept and just doing tons of different art styles with it 😄

[late blade]

that looks like some square enix cutscene on a PS4

dall-e has so much problems when it comes with laughs and teeth

Prompt was:

Create a wide, ultra-realistic image of a blonde woman with a euphoric expression, casually dressed in a tuxedo, in a restaurant setting. The scene is not a photoshoot but a casual, everyday moment. The woman should blend naturally with the restaurant's ambient lighting, which should be soft and typical of a relaxed dining environment. The light should highlight her features in a subtle, natural way, without appearing staged. The depth of field should be gentle, focusing on the woman while allowing the restaurant's interior to be softly visible, complementing the scene. The image should convey hyper-realism and spontaneity, with every element, from her attire to her expression, seamlessly integrated into the setting.

[final compass]

would be interesting to test if the different emphasis on realism have any impact or if it behaves the same with just realistic.

[late blade]

I was thinking that, so I just adapted the prompt lol

removed every reference to realistic, realism or similar

problem is I still have some global settings for the chat, so it's still integrating some stuff

[final compass]

yeah i feel the chat tends to carry some context over to followups even if it isn't explicitly referenced.

[late blade]

the last two were the best I got, now it's starting to degrade the concept

[final compass]

haha, there is some hopelessness in getting a magically good generation at times and trying the best you can to spawn a whole series based on that one image. sometimes it works beautifully, other times not.

[late blade]

I wish I could predict memory tokens in chat

afaik, as I understood it, each prompt has a limit of tokens, some used for memory/recalling details, and some for out inputs

[abstract crest]

anyone noticing incompetency in DALL-E image generation today? Reminds me of GPT incompetency when they try to release something. I'm guessing they're modifying the code for that 'lazy' update. I read DALL-E needs to be updated

[late blade]

You'll have to deliver more information

[onyx ridge]

No, I don't think it's lazy at all. When I ask for 1792 x 1024, not 1500 x 900. What do you mean, lazy?

[late blade]

I still don't get it when it comes to what "incompetency" means

I think people still hope that they have a vision in their head and it will work as people imagen it at the first try

[abstract crest]

I didn't say it was lazy, I said it seemed incompetent today.

[abstract crest]

AT least you properly read my comment and noted the key thing I was referring to. Simply put, incompetent is just generating images badly today as opposed to, say, last week.

Last week I got better results. Today, great frustration

[onyx ridge]

Prompts?

[abstract crest]

So just wanted to see if anyone else noticed it compared to the usual results. I saw the new version of turbo was released

[onyx ridge]

Before/After?

[onyx ridge]

No.

[abstract crest]

I'm sorry, do you have a proper question?

[onyx ridge]

1. Which prompts are you using?
2. What do you want from the model?
3. What's it not doing?
4. What were your successful prompts?
5. What was the difference in output?

[abstract crest]

I'm not that interested in diagnosis of what may be wrong with my image generation. Just asking if anyone noticed something out of the ordinary

it's a custom GPT anyway, so

[onyx ridge]

So you didn't have proper questions. Got it.

[abstract crest]

but I use the same prompts almost always which is why I can see difference when it doesn't behave as usual

[abstract crest]

I don't think you got anything, but sure 😄

[onyx ridge]

It seems like you got no images. Do I got you?

[abstract crest]

it's ok, I'll just experiment with different prompts. It's more of a 'free prompt' which I assume why GPT can generate strange things sometimes. When I say 'free prompt' I mean about 300 words in instructions, plus about 100 words in a prompt. I set quite a lot of criteria, but within that criteria, GPT can generate whatever it wants. It's for blog post thumbnail images so I'm guessing as I go through all the topics, it has to learn what is acceptable for me. I'm almost positive this happens because a new DALL-E update is coming and I am affected, but it could be random. So thanks for the offer to review my images and prompts, but I don't think there's much to be gained 🙂

[onyx ridge]

Regrets for not blindly echoing your experience and trying to diagnose the issue. Good luck.

[abstract crest]

there's always next time

[onyx ridge]

So you use a long prompt with a history of variable results, but when you got variable results today you blamed it on the model change?

[thick smelt]

hello guys enjoy my upgrade of the woman with tuxedo 🩷 🤵‍♀️

[late blade]

It's also the usual pattern, people say they have a problem, they don't specify and then there's questioning from the regulars

[fickle ravine]

Wow, really nice one

[fickle ravine]

How was I supposed to know that, they are too paranoid about their content policy, that system has an IQ of a frog

[onyx ridge]

Looks a little bit like Brec Bassinger.

[fickle ravine]

Today I generated a few pictues, I don´t want to overshadow others in gallery so I will just post it here

[late blade]

if there's something to discuss, do it, we are all ears (eyes)

[thick smelt]

thanks i guess

[fickle ravine]

Not really, I just wanted to post them here

I am just a little mad because of how messed up that content policy is

[final compass]

@abstract crest Prompts need to be precise for precise results. But I can relate to experiencing different degree of 'problems in generating images' as content policy has changed somewhat over the course of time. There have been some instances in the past that i was generating a series of images using pretty much same concept and style and sometimes the results were just for no apparent reason way higher quality. This is hard to quantify if it was luck, if it was some changes done to dall-e, if there is some sort of budget of 'how expensive each generation is computing wise.'

[thick smelt]

and of course the legendary gecko that we love here in #images-discussions , he has to join the tux party

[late blade]

blame GPT4V

[fickle ravine]

Waiting for the day when rage will be the daily theme, this is perfect

[thick smelt]

GPT 4 is the 🐐

[fickle ravine]

I wonder how does AI mess up the German flag like this, look at it closely and you will see the orange color between red and yellow

Also the text

[late blade]

ask for a flag alone, no PC and test it out

[tall mason]

I got a notification from this convo for some reason. In any case @abstract crest, long convos with GPT start to lose coherence and you may need to start a new one. Also note, what you tell to GPT is not the prompt given to Dall-e. GPT will often add superfluous details, especially after long conversations. We'd need prompts and images to say more.

[fickle ravine]

I have used one convo for a LONG time, and then it just broke down and I had to start a new one, it was not on ChatGPT4 tho, it was just a standard model

[late blade]

this happens because the GPT "Memory" is a number of tokens that are used for the whole prompt, so the moment we start using more tokens, the diffuse the information gets

[thick smelt]

oh yeah i have the same problem with the japanese language keep messing up, it's not the word i prompted but here's an example of it, it says "Rakyu"

good

[fickle ravine]

looks actually pretty realistic for an AI creation

[late blade]

her left eye has some weird blue shading

[dim cradle]

/video

[fickle ravine]

When I said "You didn´t answer my questions." it suddenly worked

The AI really needs a brain

Really???

[late blade]

really? what?

[fickle ravine]

It really dared me to ask about which answer is better after changing nothing

It just changed can´t to cannot

[late blade]

GPT doesn't compare the answers, it just shows them to you

both outcomes had the same answer

[deft musk]

For the interested, found a second way to help improve Dall-E 3's spelling: #1201404302921580646 message

If the quoted text is the start of the prompt, seems to do much better. Example:

"Caution: Time Travelers Crossing" sign in a chaotic urban street scene. The sign is prominent in the foreground. In the background, various anachronistic elements are present: a medieval knight on a horse, a futuristic hovercar, a person in 1920s attire, and a dinosaur. The street is bustling with confused modern-day pedestrians and vehicles, adding to the comical chaos of time-traveling entities appearing in a contemporary city setting.

[fickle ravine]

Incomporating foals into every type of picture is fun

[open trench]

morning all 🙂

[thick smelt]

good morning nezhno chan!

[fickle ravine]

I am sure that if I told AI to generate someone who died it would refuse

Oh yes this way you don´t use that one word

[verbal sequoia]

<:book_icon:1171408210398289941> `` Rule 2 `` Keep it G-rated.

Do not share or display content that is not G-rated, including profane or vulgar language, via any means. NSFW or violent/disturbing content is strictly prohibited.

[fickle ravine]

I got a warning today for the same problem

The rules are too strict, but it is what it is

[late blade]

It's not strict, you have to adhere to the rules

[hot rain]

G is a lower rating than modern kids movies like Paw Patrol and is pretty comically strict lol, but I digress.

[late blade]

yes, we talked about this many times

[fickle ravine]

Yes, too strict, once chatGPT refused to write a story about rescuing foals from a foal torture facility because it promotes harm or violence towards animals

[late blade]

for now, #1070006151938314300 and hope there are changes

[fickle ravine]

Like sorry but that is just hilarious

How does saving foals from violence promote violence?

[hot rain]

I asked it about a hypothetical fictional fight between a grizzly bear and silverback gorilla, and it admonished me as though I had the animals in cages right now and was about to pit them against each other after it answered lmao.

[fickle ravine]

I love trying to include a foal in every picture I generate

I hope the new theme will have as much potential as this one

[thick smelt]

dalle has a problem on the japanese word spelling, my pikachu sticker has "Rakyu" for some reason

[fickle ravine]

Not just japanese

[thick smelt]

oh my bad, english is so good at dall e in my opinion except for few times

[fickle ravine]

The more in the middle of the text in my picture you are the worse it gets

[thick smelt]

yeppers longer text are a hassle to dalle 3

[fickle ravine]

Idk if this theme has a lot of potential, this one is pretty much balancing

[dim cradle]

@onyx ridge thanks for creating the gallery -- sorry about all the hassle -- looks awesome, dropped some images in there #1201430692509261886 message

[onyx ridge]

Thanks for the feedback. At least the mods were helpful. Still, that automation of deleting posts for just a handful of legitimate mentions is very disheartening.

[fickle ravine]

Wow this is VERY realistic, by far the most realistic thing I have ever seen AI make

[onyx ridge]

I'll continue to try to engage with the Gallery format, but if this happens again for some other arbitrary reason, I'll be boycotting it entirely.

It is objectively not a good format, especially since there are no rules or instructions specifying how to handle accreditation of collaborative work.

That said, I aim to follow the rules, even when the rules are absolute #failsauce.

[fickle ravine]

Yeah the same as the content policy

Like, I understand it marked halter as a violation. But in the context of the chat, it was crystal clear I was talking about foal training

[onyx ridge]

This is different. The content policy, while focused on English, can at least be researched and understood, given the incentive.

There is no posting for the rule that I can't mention more than two people in a post.

The rule exists but it's not recorded anywhere, it's an arbitrary undocumented setting on the local bot.

I couldn't have researched this.

In fact, it's very different. I was penalized by a faceless system for ethically crediting collaborators. That's far different than the machine declining to do work that could produce harmful imagery.

There's a similarity, of course. A bot is responsible in both cases. But one is well-documented and avoidable. The other is not documented and only avoidable if you inherently don't care about giving credit where it's due.

[dim cradle]

Seems we ought to be able to mention multiple users in the same message -- a few mentions isn't spam.

[onyx ridge]

I won't rant about it further, but since it came up I figured I'd give my thoughts on it again.

Sorry for the ranting.

[dim cradle]

Like raise the limit from 2-3 to 5-6 and the problem is probably solved.

[tall mason]

I wonder if that's an issue with discord. Like, either multiple people can be messaged at once, or only individuals.

[onyx ridge]

The message last night was that it's a bot setting.

[dim cradle]

i get the impression it's the way the spam detection is currently configured, it's rather too strict perhaps

[onyx ridge]

Frankly, while I'm a unique use case (artistically talented, visually disabled), it's almost impossible for me to believe that I'm the first person to try to credit users for collaboration through mentions. How long has the server been operating and with how many tens of thousands of users?

[fickle ravine]

I wonder why does the mod team not even care

You can use modmail to ask eventually

[tall mason]

About a year and a half I believe. A lot of users weren't all that active, except for when new models were released.

[onyx ridge]

No they do care. They were pretty prompt to fix it to my liking, actually. While they asked clarifying questions, they understood I didn't want a weird gallery up with a deleted first post. They understood my concern that it looked like I posted something weird or unethical, when in fact my message was deleted for following ethical guidelines.

[fickle ravine]

oh okay then

[onyx ridge]

I just don't want to cultivate an "us" vs "them" sentiment, Lukáš. Instead, I want to draw attention to the systemic flaw, and hopefully unite the mods, council, volunteers, and community around the principle that ethical credits to other users should be encouraged, not discouraged, by the systems we share.

Maybe the bot can be given an exception to the Gallery and any other "collaborative" channels, for example.

Maybe users can be given a quota of @mentions to ensure this isn't abused.

It could even be gamified, encouraging collaboration by unlocking @mention budget by tier, ensuring that it's budgeted to collaborators focused on community.

[tall mason]

The trouble there is the bot, like the starboard bot is, may be third party.

[onyx ridge]

It doesn't have to be set up that way, of course, but I understand blanket @mention power might be too much.

[onyx ridge]

I'm pretty sure, but not positive, that some of the admins can control parameters of the bot. Depending on its design, it may allow exceptions.

Clever programmers would include exceptions, anyway.

[fickle ravine]

I think they can, but the programming is not that easy

[tall mason]

No doubt. But if not, their options are request stuff from the devs, find another, or make their own.

[onyx ridge]

It seems nothing worth doing ever is, friend.

[onyx ridge]

When ethical posting leads to system failure, transparency is in order, I think. That's not a call to arms or anything, merely an insight.

[fickle ravine]

I hate that 30 mins cooldown in daily theme, I just generated 3 perfect images, I am going to post them at once

[onyx ridge]

If the bot has those features, I'd encourage the mods and council to share a plan to avoid penalizing ethical accreditation mentions.

If it doesn't, it behooves them to clarify the rules as they're programmed and make them available with the rest of the server rules, to promote trust and equity.

This doesn't have to be couched in mystery and secrecy.

There's no paradox at play here due to a G-rating requirement, either. Unlike poor Lukáš, I can talk about ethics with G-rated language ad-infinitum.

His issue is more nuanced.

The rules forbid me from flat-out explaining what would be obvious absent the language barrier.

It's unfortunate, but those are the rules. Some language is associated with some acts we can't describe here.

But ethics are easy. Transparency promotes positive resolutions. Ethics makes these problems simple.

Craftsmanship and passion. Today's theme's a tough one.

[fickle ravine]

The G-rating is too strict

Both here and in ChatGPT itself

Like is it so hard for the system to recognize I am talking about halters for foals from the context of the message?

[onyx ridge]

Lukáš, if you read my posts above though, there is the best clue I can possibly give on this server as to what the issue is that you're encountering.

[fickle ravine]

But yeah, the words with more meanings are tricky

[onyx ridge]

Yes.

I regret I can't just give a proper lesson of do's and don'ts, and why's and why not's. But those are the rules.

I'm able but I may not.

[fickle ravine]

The rules are just too strict, they have literally no exceptions

[onyx ridge]

Frankly, it's uncomfortable skirting the topic. That's not your fault, but it is.

[fickle ravine]

Like censorship for that 3 letters word you use when you are confused, you know what I mean

That is just not neccessary to say the least

[onyx ridge]

I do. Talk to ChatGPT like a kindergarten teacher to a class of children, but with care to avoid adverbs.

There are a few you can use, sparingly.

Such as sparingly.

[fickle ravine]

Yeah I sometimes also do it like that

[onyx ridge]

But even that drastically depends on the context.

[fickle ravine]

The G-rating rules are just ridicilous

[fickle ravine]

In the situation of the halter it just fully ignored the context

[onyx ridge]

I want to provide perspective. It's awkward but I think I can do so with sensitivity.

Not all users are at the same place, socially, culturally, cognitively.

In order to provide an environment that's safe for people with social, cultural, or cognitive challenges to interfacing with the idea-space of the world at-large, OpenAI has to make some compromises.

When you consider the scope of over 8 billion potential users and their broad sociocultural and cognitive conditions, this challenge is monumental.

In fact, if you think about the extreme differences between some of the above-mentioned conditions, it's incredible that they can achieve anything approaching that goal at all.

[fickle ravine]

Yeah, it is funny how they trained it to be culturally sensitive, but yet it can provide recipe for a beef but not for a foal steak

Beef can be sensitive to people in India, same as foal steaks can be sensitive to people in America

It clearly thinks American culture is superior, and that is just bad

This is used just as an example

[onyx ridge]

Its cultural training is built from data disproportionately clustered around San FranSicso, CA. That's an unavoidable bias.

To be candid, and I man no disrespect, I had to use the internet to search this dietary preference world-wide to understand its prevalence.

[fickle ravine]

Beef is generally considered bad in India, with a few cultural exceptions

[onyx ridge]

Over 8B people and a propensity toward cultural isolation makes perspective on diet particularly challenging.

[fickle ravine]

I am not sure about any specific one, but I am sure a lot of cultures consider foal meat as good

[onyx ridge]

But as I said, the model's cultural training, while quite diverse, is still disproportionately clustered around SF CA.

[onyx ridge]

My research indicated that nearly 1/3 of the world do.

I mean no insult when I say, "I had no idea."

[fickle ravine]

How could that be an insult

[onyx ridge]

People are weird about food. I just wanted to clarify.

[fickle ravine]

yes I understand

[onyx ridge]

I think that's culturally universal, too. People, in general, are weird about food.

[fickle ravine]

For the mod who thinks we had a discussion about religions: No, we were just discussing if and how AI respects different cultures

[late blade]

can you guys please tone it down with the ranting? it's been a while now, ground yourselves, breather 30 seconds, get a drink and back to creative mode please

[feral minnow]

If you want to create things that happily go to PG-13 places, the answer is Bing. Same groundbreaking model and not subject to the OpenAI ethics guidelines (just Bing’s much more lax content restrictions)

I happily use both.

[onyx ridge]

Apologies. Some of the cultural discussion aimed to facilitate creative dialogue through understanding the breadth of users.

[fickle ravine]

Sorry, we were just discussing AI´s respect for different cultures

[onyx ridge]

I'm struggling to decide on an image style for today's theme.

[fickle ravine]

You can do several styles

My style are ralistic images, (almost) every single one involving a foal even if the topic is not the best for it

[late blade]

hmmm, how about instead of LED pointillism, do something like coin pointillism? or cotton pointillism? try new textures?

and it could be a mechanic or engineer crafting Lexideck

or something completly different stepping out of the comfort zone

[fickle ravine]

What is your favorite style of pictures?

[late blade]

it comes and goes, depends on the mood and if I'm hungry or not

[fickle ravine]

And some mayonese

GPT4 is pretty good at generating food not gonna lie

[late blade]

nice, but already had dinner, if anything something sweet would be nice

[fickle ravine]

Even for me, who is not a steak person and hates toamtoes this looks really good

[late blade]

I'm an omnivore

[fickle ravine]

Yeah me mostly too, I just hate tomatoes and strawberries

[onyx ridge]

@late blade @dim cradle
"Mirai Kanketsu-shugi" (未来簡潔主義), or "Future Minimalism," aptly encapsulates the essence of Retrofuturistic Minimalism within a Japanese context. Drawing parallels with traditional Japanese aesthetics such as 'Ma' (間) - emphasizing negative space and simplicity - and 'Wabi-Sabi' - finding beauty in imperfection and transience - this style resonates deeply with the Japanese artistic heritage. The sleek lines and geometric forms of Mirai Kanketsu-shugi align with the clean and functional aspects of 'Zen' design, while its futuristic elements echo the imaginative and often whimsical nature found in contemporary Japanese pop culture and media, such as anime and manga. This fusion not only reflects a blend of eras but also a synthesis of deeply ingrained Japanese artistic principles with a modern, forward-looking vision.

[grizzled loom]

good day my people

[fickle ravine]

Why do I find that cute 😭

[late blade]

derivative from this: #daily-theme message ?

[onyx ridge]

No, mashed up from a combination of "Ma" and retro-futurism.

I hadn't seen that post yet.

Correction, I had. I didn't associate the two.

Other than it's inspired by the daily theme.

Theme > Style

Convergent evolution.

[fickle ravine]

Members of football club FC Twente Enschede visiting a shelter for foals from "not ideal" circumstances, generated for a friend

[late blade]

ahaa, I was just wondering because of Wabi-Sabi

[onyx ridge]

The AI wrote it up.

I just mashed up simple Minimalism and Retrofuturism.

[late blade]

I'll post my image in today's theme, got something good to show you

[onyx ridge]

Then I asked for a Japanese parallel and style definition in a single entry, derivative of Shon.

[fickle ravine]

God why is all of it so cute

Also the blue skies in the meadow remind me of like 2010-2014, when I was a little kid

[quartz vale]

[fickle ravine]

@late blade I like your newest one

She is actually pretty, just dirty

[grizzled loom]

im working on total cohesive beavior of ChatGPT and Dalle together, again. I had itfixed before, but updates introduces change every time breaking the work. About done with GPT "Five" lmao

[late blade]

mechanics are always with smudges

[fickle ravine]

true, except for the one I just uploaded

But this is more of an ingeneer than a mechanic

I will ALWAYS find a way to implement a foal in the themes

[late blade]

next in queue is what @onyx ridge proposed, with my own twist

[fickle ravine]

Oh my god that is just sooo cuteee

The faceee

My pictures can never compete with this, but I am still pretty happy about them

[late blade]

hmmm, interesting, want a nobel in physics with that? and a side order of fields medal for math? just with a side dish of fries?

[grizzled loom]

Usually I prefer a salary or business partners for work, this will be on the house.

[onyx ridge]

Now it's on! 😄

[fickle ravine]

Why is there that Plexiglas 💀

It looks like it

Still cute tho

[grizzled loom]

a nobel in physics would be for writing the grand unified field formula. with full proofs.

[fickle ravine]

Do you allow me to download this into my PC?

[grizzled loom]

the directive? i will hvae it in the gpt, but the directive is meant for the team to fix the system

[late blade]

or erasing misshaps for a 100%

[fickle ravine]

#daily-theme message

Again a definition of underrated

[onyx ridge]

I really wasn't trolling about the brighness of the images. Without significantly tweaking my display (breaking calibration) I can't see the details.

[late blade]

I know how it feels, I got light sensitive due to a hypotonic pupil, sometimes it's hard to see certain colors

[onyx ridge]

I get to pick between accurate color and blindness to brightness beyond a certain level, or inaccurate color that I can see in its entirety.

My pupils don't contract at all.

They exist at about 83% dilation all the time.

HDR screens = #hardpass.

[onyx ridge]

Beautiful!

[fickle ravine]

I love how AI refuses to generate a foal tied to a fence yet it generates a foal in unnatural position like this

Is that a chicken or some rare exotic chilli pepper?

Still looks really cute

[grizzled loom]

its an ethnic cusine. this one is indian food

[fickle ravine]

It reminded me of those extra spicy exotic chilli peppers that appear in the news

[late blade]

you showed this yesterday, what tool is this?

[onyx ridge]

@late blade I think your recent post qualifies as "Mirai Kanketsu-shugi" (未来簡潔主義). I will probably make a gallery later for the evolution and exploration of the style. I hope you'll add these lovely images. They're lit such that I can see them well! 😄

[late blade]

these could have been perfect for the last theme

[onyx ridge]

Yes.

There's a good deal of overlap without purpose in the subject.

Which I'm not averse to. I like that they're mirrors, too. No matter how hard it is for me to stare at this style.

It's beautiful, just brutal on my retinas!

[late blade]

I'll check if I can do something with a darker color palette

[onyx ridge]

Nevertheless, reliant in resilience and knowing it's not as bright as #outdoors, I'll continue the exploration. It's fun, if challenging.

[onyx ridge]

I was going to move to "moody" lighting and palette next.

With the same style.

[fickle ravine]

I swear I will include a foal in the pictures even when the theme is a futuristic city

[dense mesa]

well that's comfyui, and then I'm using one of the nodes I created for it. the interface is a lot more versatile than most people realize, most of the users might not use dall-e 3 with it. but it's actually quite convenient

[onyx ridge]

What about a massive equine cloning facility and training ranch in a megastructure set in 2099?

[late blade]

cool, I've never worked with comfyui before, I might give it a try

[fickle ravine]

Yeah this is also a good idea

[onyx ridge]

I try to be helpfoal. 😄

[dense mesa]

what's great is you can then send it right into all sorts of post processing. so have it auto-upscale as soon as it's finished rendering. working on some other tools more geared towards dalle3. but a lot of the stuff that already exists is pretty interchangeable

[fickle ravine]

I am foal of gratitute for your support

[onyx ridge]

Be carefoal with the puns.

[fickle ravine]

Do I foaly understand they can violate rules? Those rules are not usefoal

[dense mesa]

if you have any questions let me know. I know a little bit. and then if you want to use that dall-e 3 node I have my github linked in my profile. also listed in the comfyui manager

[late blade]

excellent, I'll check your github and see your inner workings

[dense mesa]

well most of those nodes are actually pre-dalle 3. but then when I went to use the API it just made sense to me to utilize that format/interface. working on a couple more for openai stuff. but not quite there yet

[fickle ravine]

That tree looks so cool, I love the shade of green

[wanton moon]

Thank you, I wanted the tree even tho it is kind of a holographic sculpture, to come to life in the bleak surroundings, I am glad it turned out that way

[dense mesa]

strange output there

[onyx ridge]

Thank you for making my day...

[fickle ravine]

cursed

[wanton moon]

lool David Bowie

[ripe scroll]

Ah, found the book i was interested in attempting to enter into one system or another. Johann Wolfgang von Goethe's Versuch die Metamorphose der Pflanzen zu erklären, or Metamorphosis of Plants. Can't remember exactly what about this book was interesting previously, have to peruse it again, but all the diagrams and such, collecting hand made material from effort of representing natural processes

[ripe scroll]

I'm more David Belle heh

[wanton moon]

Calmm your HHorses People, All HHorses are available for Adoppion

[fickle ravine]

Wait why does one of the foals in my newest daily theme picture have 5 legs

[late blade]

I'm getting really accurate life images instead of plastic

[fickle ravine]

And why did I not notice until now

[wanton moon]

my headcannon is now that the horse was grown on a genetic horse farm and a mutation happened

[fickle ravine]

Maybe due to that it was supposed to be destroyed and then it was rescued

Lore revealed

[wanton moon]

so does that mean that this horse is now 1.25 Horse Power?

[fickle ravine]

Probably

Pregnant horse is 2 horse power

@verbal stream to be completely honest, your picture is one of the best things I have ever seen. It is so cute

[verbal stream]

[fickle ravine]

It definitely deserves votes more than my foal playground

[late blade]

@dim cradle is alive

[late blade]

ok, i'm out of classic ideas, time to spice things up

[onyx ridge]

@dim cradle #daily-theme message

I hope you approve of this style!

[dim cradle]

Mirai Kanketsu-shugi + today's theme

[dim cradle]

Love the brightness

[onyx ridge]

I toned it down over each evolution.

The most recent used "moody ambient lighting, and a cool gradient pastel palette."

[onyx ridge]

Thanks!

[wanton moon]

for the lols

[dim cradle]

the style results in some cool anthropomorphic AI

[onyx ridge]

That's what I have focused on for the subject!

[dim cradle]

i saw yours in the factory scene and tried it on a closeup

[onyx ridge]

It's so good

[wanton moon]

first person perspective of a person seeing the full color spectrum including infrared and ultraviolet, magnetic fields in the sky and visible vibrations of sounds nearby

[dim cradle]

@open trench dropped a few in City Invasion #1201572168241250435 message Thanks for the helpful specs.

[deft musk]

I like your ideas and suggestions here. I actually recommend maybe think of a list of ideas, and present them to modmail.

It's not about preventing giving credit, that's an unintended side effect.

It's not about preventing people from getting the attention of folks who want to know, that's also a unintended side effect.

Apparently the mods went through a kind of overwhelming 'attack' here, that made the bot setting 'the answer that made it managable'.

Maybe things are better now, and/or maybe there's a way that's better than what we have now.

I ran into my own experience with the @ limit when I did too.

To my thinking, the sum total of here is very positive, and all the good + all the bad is more than enjoyable. But the place can be improved. I love seeing your ideas to make it more workable and better, and still keep some control. I don't know enough about Discord to even know if they're implimentable, but I do bet, you can put your ideas together and directly offer them to the mods. You might even want to ask them, if the ideas are not useful, to please let you know what some of the problems are, so maybe you can help brainstorm solutions they can use that we can all enjoy better than what the current methods are

[boreal gate]

Whoever suggested using "typeface" thanks

[dim cradle]

i believe that was @deft musk

[deft musk]

You're welcome! I see it still struggles, but that's quite a bit of text there!

[boreal gate]

Yea it's a whole process

[wanton moon]

epitome of feline grandeur – a majestic big cat, a wondrous fusion of lions, tigers, leopards, jaguars, lynxes, cougars, pumas, cheetahs, and the ethereal Siberian white tiger. This resplendent creature, draped in a luminous white coat with a mesmerizing array of patterns, reclines gracefully in a diverse habitat. A harmonious symbol of the wild's unity, strength, and beauty, embodying the essence of an extraordinary blend of big and small cat breeds

I like big cats and I cannot lie

[open trench]

They look great!

[dim cradle]

thanks, just some initial results from your specs, just to get a feel, need to refine it and focus on a scene

one automated variation indicated a military response.

[open trench]

neat

I'll be very curious to see where it goes

[dim cradle]

maybe the citizens will fight back also lol

[open trench]

or a superhero, even.

[dim cradle]

all of the above

[open trench]

lol yup

[dim cradle]

haha

[open trench]

i hope more people participate too. that would be cool to see.

[open trench]

I added the specs to help guide, but I hope they aren't too restrictive

[dim cradle]

just the right amount of creative direction ha

[open trench]

nice

[full scroll]

I have been unable to buy dalle-2 credit in any browser. Firefox, Edge, Chromium. Do they not want people to use it? I have an account with a payment method on file, they have no problem billing me for gpt api usage.

[dim cradle]

Potential superhero in @onyx ridge 's style...

The reimagined superhero in the 'Mirai Kanketsu-shugi' or 'Future Minimalism' style has been successfully created. This portrayal blends the sleek lines and geometric forms of futuristic minimalism with traditional Japanese aesthetics, presenting the superhero as a figure embodying modernity, minimalism, and traditional Japanese artistry, set against the backdrop of the city invasion.

[open trench]

Nice

[onyx ridge]

That's awesome

Thanks for continuing to collaborate!

[full scroll]

Is there any solution for the white screen after attempting to buy credits (any browser) or is it just hopelessly broken with no acknowledgement or fix in sight?

[onyx ridge]

Hiya!

Anybody heard of Blackout Diffusion?

Fromthe arxiv paper:

Typical generative diffusion models rely on a Gaussian diffusion process for training the backward transformations, which can then be used to generate samples from Gaussian noise. However, real world data often takes place in discrete-state spaces, including many scientific applications. Here, we develop a theoretical formulation for arbitrary discrete-state Markov processes in the forward diffusion process using exact (as opposed to variational) analysis. We relate the theory to the existing continuous-state Gaussian diffusion as well as other approaches to discrete diffusion, and identify the corresponding reverse-time stochastic process and score function in the continuous-time setting, and the reverse-time mapping in the discrete-time setting. As an example of this framework, we introduce “Blackout Diffusion”, which learns to produce samples from an empty image instead of from noise. Numerical experiments on the CIFAR-10, Binarized MNIST, and CelebA datasets confirm the feasibility of our approach. Generalizing from specific (Gaussian) forward processes to discrete-state processes without a variational approximation sheds light on how to interpret diffusion models, which we discuss.

[onyx ridge]

@dim cradle Robots serenely meditating in Mirai Kanketsu-shugi style!

/image 16:9 an anthropomorphic {MasculineOrFenimineGPTChooses} robot in Mirai Kanketsu-shugi style and hyper-realistic rendering in a zen-like environment with soft, ambient lighting, including irridescent wisps of incense smoke in the air and pseudocolor LED accent lighting.```
This is how I did it. I used my GPT, but I really think DALL-E 3 GPT will do just as well.

Notice the use of an open variable in the prompt. This way, regenerate produces a (roughly) 50/50 split between masculine and feminine bots.

These are the two subsequent regeneratrions. The open variable continues to do its job, ensuring our robot factory puts out a proportional and representative assortment of robots. In future prompts I could add other options too, I think. But for simplicity of demonstration, this is adequate.

This demonstrates that not only can open variables produced varied outputs that mimic improvisation, but that they can also encode instructions for the AI to follow.

You could write the whole prompt as an output template with open variables encoding instructions for evaluating them. It looks weird, but it works.

[formal osprey]

is anyone facing a lot of server errors with dalle3 right now? nevermind, it's a chatGPT issue

[onyx ridge]

Sweet application!

[fickle ravine]

So we are sharing robots now?

[fickle ravine]

This one looks so good

[wanton moon]

the text on the display remind me of this

[wanton moon]

Thank you

[full scroll]

Yep, this service is just hopelessly broken. So many people have this issue with any browser they try. I no longer trust them with my payment info after performance like this.

[fickle ravine]

Yeah, I asked it to explain its older answer to "not so appropriate" message using a picture

It made this

What is actually the worst theme to implement a foal?

[fickle ravine]

This theme is not that bad actually, definitely it has less potential than farms or rural, animals or horses directly tho

[fickle ravine]

Can I post a man sitting on a toilet while still wearing everything (so no adult content), apparently "fixing" the toilet in daily theme, or is that against the rules? Just asking, Ican´t be sure about this

[late blade]

For this kind of questions you should contact OAI directly. We in the community don't have information regarding this kind of errors or precedures.

[grizzled loom]

man, its so frustrating when a scene comes together and somethign so evident is off like this figure who is supposed to be holding the umbrella. even the hand posture is right.so sad. great picture if that error hadent happened. no excellent way to edit that manually for afix either. sighs

test 2 isnt horrible, but test 1 really nailed colors and look

test 3 is fairly good but some of the umbrella lost cohesive structure and has dripping off elements from the umbrella, again though not horrible.

slaps umbrella into actors hand "Listen, Cindy, I know you want recogition as a model, but You'll stil lactually have to hold the umbrella. Please leave the artistic interpretation to the artist. Your job is to be a malleable form." "do we really pay this girl 14.50 an hour? makeup! makeup!! lets goo folks! money is burnin'!" 🤣

[final compass]

Any character holding anything is surprisingly problematic. Did the blacksmith image for the daily theme and damn there were so many creative ways to hold hammer incorrectly.

[grizzled loom]

yeah, ive seen it be problematic for holding a few things; it leads me to believe there were simply no images in training data that show the nuances of griping various tools. how often is any human thinking, "sure am glad i took picturs of all these hands picking up and holding these items with all of these various grips. its a veritable gold mine."😂

[vapid elk]

what is the current image limit for dalle on chatgpt?

[wispy storm]

Last time I know it was 200 pic /24h

[vapid elk]

thx

[fair crescent]

What if there is a mention that doesn't notify people? That way you prevent the whole 'is this spam' question. It isn't ideal, but at least you can credit.

[vapid elk]

true

[vapid elk]

what is the problem with mentions?

ahh, I see.. you tried to mention a bunch of people at the same time and the bot didn't liked it

this might be changed in the future, for now, it is a patch for a simple problem: spam bots

it isn't uncommon, in fact, it is trivial to code an automation to spam links and mention as many users as possible before the staff can ban the accounts... and this discord in particular is often a target for this kind of action, so.. yea.. sadly we wave to be over zelous with that

but that is a valid point, accounts that are here for quite some time could have some exemptions, it is a nice idea

[onyx ridge]

That's a good idea!

[onyx ridge]

Three people.

Crediting collaborators.

[onyx ridge]

Wouldn't you prefer to be notified of credit for your work though, if you're collaborating?

[fickle ravine]

@stray coral I just love that ghost you posted, that is just so cute

I generated this, it looks pretty normal at the first sight, but then you have a more close look on the foals and it is not that normal

[fickle ravine]

Biscorsse 💀

[hearty ether]

If I wanna keep the same style and look for images but my chat is too long

Is there a way to transfer seed and image id across chats

[dim cradle]

my impression is that currently they're only supported in the same session

[hearty ether]

But my session is too long 😭

And I finally had a style I liked

[deft musk]

We can now copy the image and paste it into a new conversation, and get the model to continue the theme. My experiments with this, only a couple, worked really well, so might be worth a try, just upload 1-2 images you really like that showcase that style?

As a random example one way to do this, I grabbed the recent picture above for an example to compare, and showcase a iteration of comparative improvement.

And that's with an image I don't know the prompt of. You have your words from your other conversation, I hope, so you can instead/as well bring that over, the actual prompts.

[fickle ravine]

You could have asked me, it was a Bulgarian man with a foal next to him flexing his new gaming PC on Discord

[thick smelt]

dall e as a mlp from bing

[deft musk]

Of course. The goal of the example was to showcase the options Stealth has to migrate their work from one conversation to a new one, and that you don't even necessarily need even the prompt.

[fickle ravine]

Did you confuse the words goal and foal?

[dim cradle]

New beginnings.

[deft musk]

No, but that's an awesome question and use of language for a non-native speaker especially! Love it!

[fickle ravine]

Too much sunlight, otherwise this is a really cool picture!

[fickle ravine]

Can I download it and keep it for my cousin, who loves unicorns?

[dim cradle]

the sun is for optimism 🙂

[fickle ravine]

Yeah I get it, I just said my opinion

[dim cradle]

optimism, nighttime, foals

[fickle ravine]

really nice, this can be used when serenity is the theme again

[dim cradle]

generally speaking, i've always heard that an artist's growth depends on honest and accurate feedback from others, so I appreciate it and encourage it. Some of it is subjective, so I might respond with clarifications, though it doesn't necessarily negate your insights -- to clarify, my concentration is less on being an artist and more on the engineering; the sunlight was part of a test of the model's capabilities, the composition of the scene was a secondary concern:

Advanced imaging features to be utilized include ultraHDQuality, superHyperrealism, advancedLightingEffects, highDynamicRangeImaging, dynamicTextures, intricateDetailing of the apartment, the developer, and the cat, a vibrantYetNaturalColorPalette, atmosphericEffects, depthOfField to emphasize the developer's new beginning, layeredComposition, microTextureDetailing, perspectiveAccuracy, and environmentalInteraction. The goal is to create an image that not only tells the story of triumph over adversity but also celebrates the joy of new beginnings and the comfort of having a place to call home.

[hot rain]

Battle between a knight and samurai.

[fickle ravine]

Actually the recreation is not that bad, it just somehow changed his nationality from Bulgarian to Italian

How did Klement Gottwald become some Roman emperor according to ChatGPT 💀

AI usually has problems with text and specific people

I will better not share the picture of Klement Gottwald here, as by the rules it could be considered "political"!

So just Google him if you want to see whatz is wrong

Oh god all of those ones in daily theme are so pretty

[fickle ravine]

Oh wait that wall above the foal 💀

[dim cradle]

Some new galleries I recommend viewing and contributing to if you are so inclined.

Experiments with off-set focus details
#1201417562672607355 message

City Invasion (Reboot): The Rise of the Techno Mage
https://discord.com/channels/974519864045756446/1201572168241250435

[late blade]

I've been doing the ones from Nezho, so far only had 2 images worth posting there

[dim cradle]

i wanted to encourage the use of galleries and share a couple more that are good candidates, but got muted. so just be sure to browse and engage if you want to.

[fickle ravine]

They really have to fix this

[late blade]

if someone else doesn't fix it, be proactive and look for a solution with the tools you are given

[dim cradle]

The one by Yami on vibrancy and Darth's on the Japanese forest style -- no mentioning or sharing links.

[onyx ridge]

This is an open community. The problem will persist as long as it's a pitfall open to new and unfamiliar users.

[dim cradle]

I'm afraid so, the traps snare even the most experienced users.

[fickle ravine]

They could at least try, but I highly doubt they even care

[late blade]

we can only hope to do our best

[onyx ridge]

Agreed. But we can't always do our best for others in their moments of need.

[dim cradle]

we'll navigate these treacherous seas together lol

[fickle ravine]

Those rules are like fog. The only way to know there is a barrier is to crash into it

[late blade]

yeah, there's a limit of we can do, not only what we are willing to do, but also saying, I'm not stepping in

[hearty ether]

Yeah I’ve just completely lost the style of my images we can’t transfer across chats

And my old chat is full

[onyx ridge]

Always capture context.

Try copying a handful of chat exchanges and putting them into DALL-E 3 Custom GPT.

[hearty ether]

No luck

I don’t get why they won’t allow seed to be shared across chats

[onyx ridge]

If you do what I've explained elsewhere, by copying the context, shift+enter, shift+enter, then:

User
<Next prompt>

THEN prese send...

GPT can't tell the difference.

[hearty ether]

Not for images…

[dim cradle]

i suppose you lose your session seeds, but you can ask the gpt to output specs for the work you've been doing to transfer to a new conversation, and uploading images.

[onyx ridge]

The words that made those images are in the chat.

[hearty ether]

Not the seeds…..

[onyx ridge]

They're just words.

[hearty ether]

seeds affect it quite a bit

even with the exact same prompt

[onyx ridge]

We never had seeds, @hearty ether

[fickle ravine]

@lean turret I love the newest picture, however, it would fit better in some futuristic theme

[hearty ether]

Seed + image id

The combination works well

In a single chat at least

[onyx ridge]

okay, but in that case you need to be specifying your own seeds.

[lean turret]

i mean the pug is creating its clone so...

[onyx ridge]

If you're relying on a seed that you don't know the ID of, then when the context ends so does your seed.

However seed != style

[hearty ether]

I know….

[onyx ridge]

the style of your images was conveyed by the words, not the seed nor the gen id.

[hearty ether]

But it helps to refine it….

[onyx ridge]

Good luck, and I recommend to track important data in the future.

[hearty ether]

I have the info…

It just doesn’t work across different chats

[onyx ridge]

That's right. Because seed isn't accessible, only gen id.

And context is limited. So use the words

That's how we all do it.

[late blade]

This #daily-theme message is why apple doesn't sell luxury computers...

[hearty ether]

I try to generate a image every few texts to keep it up to date

[onyx ridge]

But I'm trying to explain the words define that style. The seed, gen id don't.

It might take a little work to nail it flawlessly, and you might never, but I've regenerated images dozens of times and gotten the same style from different gen id and seeds.

So I know it's not only possible, but actually the default way DALL-E 3 works.

[hearty ether]

Like my prompt has the style as realistic illustration, but it’s really easy to get it wrong

It either leans too far into black and white for illustration or uncanny realistic

I just like this innetween but it’s mostly rng so far

[onyx ridge]

I see. I can help you understand this. You need tighter constraints in your prompt. What kind of illustration? Cartoon? Photorealistic?

Be as specific as what you want with you rlanguage.

[hearty ether]

I mean tbh I don’t know how to describe this style

Not even the ai can recreate it

when I ask it to describe to a blind person then generate it

[onyx ridge]

Is the above image an archetypical example of the style?

[hearty ether]

Yeah

[onyx ridge]

What's your favorite image in the series?

[hearty ether]

Yeah just that one

[swift oriole]

Try Hand-Painted-Photorealistic in the prompt

[dim cradle]

@late blade nice i was thinking about an apple image for craftsmanship

[onyx ridge]

@hearty ether

This might be a start for you:

Imagine a beautiful blonde woman in pragmatic clothing and a belt-clipped survival knife, seated in the sand and stones next to a lazy creek in the evening sunlight. The image is illustrated in the Magic Realism style with a warm palette, natural ambient lighting, and candid theme.

I genuinely hope this gets you back on track, or at least close to it.

These are the images it produced:

The left image is honestly remarkably close.

[hearty ether]

Thanks but yeah this kinda highlights my issue with recreating the style without seed

It either ends up being the uncanny realistic dalle usually do like the left case

Or the weird illustration type on the right

And the realistic illustration style is a delicate goldilock zone that’s mostly luck based

[onyx ridge]

You can adjust the image by the description. I just expended very minimal effort to try to recreate your image. It's not hopeless to get back on track.

Define your own seeds.

[onyx ridge]

DALL-E is language-based if you constraint it by language, else luck.

[hearty ether]

this might do tho

its just far easier with seed and image id for control

sure you can get similar with language

but having 3 different points of control over the ai would be nice

hopefully they can allow us to cross chat

[onyx ridge]

Incredibly similar. Try my custom GPT sometime, and make /selfie images of an agent.

[hearty ether]

this way we can also share seeds with friend

[onyx ridge]

They come out the same in every instance, no seed required.

How?

I described everything in the image.

For all six agents.

So they come out the same.

[onyx ridge]

@hearty ether As you can see these are new instances with three word prompts.

[fickle ravine]

Ai really loves 5-legged foals

[onyx ridge]

From the DALL-E 3 generator with the same prompt, obviously an entirely different seed and gen id:

My only complaint is that "Manga-inspired" can create anything from a child-like face to a mature one, so I need to work on that language a bit.

However, the point is the style is consistent.

[fickle ravine]

It looks really cool

[onyx ridge]

Remarkably consistent.

[onyx ridge]

Thanks, @fickle ravine!

[fickle ravine]

It would look better with blue hair, but that is just my opinion, it is still really nice

[onyx ridge]

Subjectivity is weird. But I will keep blue in mind, if I ever add a seventh agent. 😄

[dim cradle]

yo

[fickle ravine]

This time AI made me happy

Not that bad result

[dim cradle]

nice

only i would be nervous to have my lunch with it near me haha

[fickle ravine]

Do you please have any idea why did it fail to make a statue of Klement Gottwald in this picture?

[fickle ravine]

It is one day old, that would only drink milk if you have it for lunch

[onyx ridge]

It's not allowed to depict real people. Seriously a risk for deepfakes if it did.

And a statue to a photoreal image is a very slippery slope for a language model. That might be hard to understand, but this is probably best.

Try describing Klement Gottwald.

Without invoking the name.

[fickle ravine]

Or maybe I can use his picture

[onyx ridge]

It'll fight you on that.

[fickle ravine]

When you compare it with the picture you can find on Google, this is not that bad

It got much better

[deft musk]

I get your concern. #server-rules are here if you want to read them, and to my understanding this picture is very okay within the rules.

A good 'test of thumb' - if it can be said or shown in front of the entire community - little kids, the workplace, 'generally not offensive to anyone, and generally G-content' - then it is okay to share on this Discord.

[onyx ridge]

Problem: Not all societies use cinema ratings to gatekeep content. This is a pretty relative scale, especially globally-speaking.

[fickle ravine]

How did it know it is Greek???

[onyx ridge]

Especially notewothy within this context is the struggle to understand what children's films have to do with AI content generation.

[onyx ridge]

Encyclopedic training, possibly even on encyclopedias.

[fickle ravine]

I will try with a place of my childhood that is not that much known

But people who know that island can recognize the exact place

[fickle ravine]

Do you know it by any chance?

It failed

Actually it is lensible, so if it used Google Lens it would be easier

[onyx ridge]

It doesn't have that capability though.

Just training.

Gargoyle ex Machina:

[fickle ravine]

I love the colors of the second one

[onyx ridge]

Me too.

The first one's also pretty evocative though

I like both and thought I'd share.

[fickle ravine]

Yeah, also the location of that bridge is Pag, Croatia (my childhood vacation destination)

[onyx ridge]

This is my least favorite so far, the model going in the wrong direction with the gargoyle, for my vision. But the sparks on the sides, I love.

I want two buttons. 😄

[fickle ravine]

Still it looks really good with those colors

[late blade]

Damn it I just noticed something on the post I made...

[fickle ravine]

Yeah I noticed this foal has an extra leg

[onyx ridge]

This is more like it.

[fickle ravine]

Epic

[onyx ridge]

Not yet but thanks!

[fickle ravine]

I asked AI to generate Gustáv Husák with a foal using a picture, and the result is mid

It made an old man with glasses

That is accurate, but the more detailed parts are not that accurate

[onyx ridge]

I think I finally nailed it:

You can't build Rowko's Basilisk without a Basilisk! 😄

I love it though! This is so realistic!

[late blade]

This was my prompt:

A wide image of a highly sophisticated circuit board representing bleeding-edge technology. The circuit board features a dragon intricately integrated into the circuitry, blending seamlessly with the components. The wiring and layout of the motherboard are designed in a neurographic style, resembling complex neural networks. The overall appearance should be futuristic, detailed, and embody the fusion of technology and artistic design.

[onyx ridge]

I just forgot to post mine. I'm a bad.

Imagine photorealistic oil painting in a 16:9 format, focusing on an HD Neon and LED Pointillism style. The image depicts an elaborate and mystical circuit board with a very subtle and majestic gargoyle silhouette embedded in the circuitry, facing the viewer. The artwork should blend technological and gothic elements. The gargoyle, with glowing LED eyes, should convey a 'draco ex machina' theme, subtly integrated over the electric labyrinth. The color palette should include vivid, warm tones ranging from subtle to pseudocolor, to suggest arcane power and technological sophistication. The style should emphasize a juxtaposition of ancient mythology and futuristic innovation, enhancing the modern yet mystical ambiance with dark, contrasty High-Resolution LED and Neon Pointillism effects.

I got a lot of LEDs

This is so good

[late blade]

Had to add a bit of my style to it

[onyx ridge]

I would hope so!

[hearty ether]

There is like the trend of making something more “European, French….)

But any one tried going less?

[onyx ridge]

I'm honored you liked it enough to make it your own, @late blade

[hearty ether]

Like making a French person less French 10 times

[onyx ridge]

I've not.

ooh I got an idea.

New prompt:
Imagine photorealistic dark sci-fantasy illustration in a 16:9 format, focusing on an HD Neon and LED Pointillism style. The image depicts an elaborate and mystical circuit board with a very subtle and majestic gargoyle silhouette embossed onto the circuitry, facing the viewer. The artwork should blend technological and gothic elements. The gargoyle, with glowing LED eyes, should convey a 'draco ex machina' theme, subtly integrated over the electric labyrinth. The color palette should include vivid, warm tones ranging from subtle to pseudocolor, to suggest arcane power and technological sophistication. The style should emphasize a juxtaposition of ancient mythology and futuristic innovation, enhancing the modern yet mystical ambiance with balanced but dark, contrasty High-Resolution LED and Neon Pointillism effects.

Subtle changes. Instead of embedded, I used "embossed". I defined the tone a little more, and further adjusted the lighting.