AI Art

[o00o]

22 hours ago, Rubin Farr said:

Dune imagery would be great to use 

Dune came out ok. I tried around with taking very specific commands taken from the feed and these were incredible. what also works really well is myers briggs personality types and attachment styles

dune:

 

 

portraits detailed description from user feed altered to my ideas: 

 

 

 

codependency: 

 

 

dismissive avoidant: 

 

 

 

fearful avoidant: 

 

 

anxious preoccupied 

 

 

 

 

 

codependency:

 

 

 

enfp:

 

 

enfj:

 

entp:

 

 

infj:

 

 

infp:

 

 

intj:

 

 

isfp:

 

 

intp:

 

 

esfj:

 

 

[Rubin Farr]

10 minutes ago, o00o said:

Dune came out ok. I tried around with taking very specific commands taken from the feed and these were incredible. what also works really well is myers briggs personality types and attachment styles

dune:

 

 

portraits detailed description from user feed altered to my ideas: 

 

 

 

codependency: 

 

 

dismissive avoidant: 

 

 

 

fearful avoidant: 

 

 

anxious preoccupied 

 

 

 

 

 

codependency:

 

 

 

enfp:

 

 

enfj:

 

entp:

 

 

infj:

 

 

infp:

 

 

intj:

 

 

isfp:

 

 

intp:

 

 

esfj:

 

 

Have you tried modifiers like “photorealistic” ?

[YEK]

How long does it take for the programs to crunch out these pics? I imagine a long time

[o00o]

23 minutes ago, yekker said:

How long does it take for the programs to crunch out these pics? I imagine a long time

10-30 seconds. High quality about a minute max

[Real Human Bean]

1 hour ago, yekker said:

How long does it take for the programs to crunch out these pics? I imagine a long time

DALL-E 2 was originally putting out 10 1024x1024 images in around 30 seconds nut now the number of variations has been reduced to 4 and it usually takes slightly longer.

[o00o]

Did you get any good results with DALL-E so far?

[Rubin Farr]

[o00o]

So I started experimenting further with dall-e trying to create a very particular bike brand for a friend but with a uncommon color. 
 

Midjourneybot completely failed even constructing the bike while dall-e created the particular brand with immense detail on the first try while changing the as top coat painting as prompted

 

 

[Real Human Bean]

19 hours ago, o00o said:

Did you get any good results with DALL-E so far?

Loads!
I've generated around 3500 images and like a few hundred of them.

I post a small fraction of them in this thread: 

 

[o00o]

how did you get so many credits? I have to to pay for every 15 credits

 

[Real Human Bean]

53 minutes ago, o00o said:

how did you get so many credits? I have to to pay for every 15 credits

 

When I first got access you had unlimited prompts and each generation returned 10 variations, then it was cut to 50 prompts per day with 10 variations returned, then 50 per day with 6 variations, then 50 with 4 and now the current situation of 15 per month or you have to pay (which I'm not doing as I way prefer to spend money on using fast GPUs on Google Colab to run diffusion models).

[Squee]

20 hours ago, o00o said:

Did you get any good results with DALL-E so far?

I've just gotten access to DALL-E 2 and the results are a lot less fucked up than Midjourney's but also a bit more boring. Looking up cities is a lot of fun though.

[droid]

Its all about stable diffusion now.

[Rubin Farr]

boneandsickle doing more great stuff 

[no part of it]

On 7/17/2022 at 9:07 AM, Squee said:

Make sure to post what you searched for. That's part of the fun.

Also, there s a thread already. I'll go ahead and merge them.

Sorry, it did not come up in my search!

[Squee]

17 minutes ago, no part of it said:

Sorry, it did not come up in my search!

Noooo problem!

[YEK]

Sorry guys but I forgot the name of the site that you could do a similar thing as this ai but it would come out with very abstract images that you could then tweak. Any thoughts?

[logakght]

https://metaxis.digital/Acerca

Made an octopus for my site.

[Rubin Farr]

[Rubin Farr]

[Rubin Farr]

[Rubin Farr]

 

[chim]

Yay I got into Dall-E 2! Gave it the ol' goblin pilot test.

 

 

 

 

Asked for a new avatar as well but it just can't spell Supreme...

Edited August 5, 2022 by chim

[chim]

Initial verdict... For realistic looking photos and 3D renders, Dall-E wins everything. I still feel Midjourney is more creative, possibly by having a major dataset from contemporary digital art. It has more customizable prompt functions and is fun to use. Look at o00o's midjourney pictures and how imaginative they are, that's difficult in Dall-E. How to even manipulate Dall-E prompts is a dark art and mostly guesswork. There are no specific tags like the :: separators or --no command in midjourney, you just have to try various ways at constructing natural English sentences. "A photo of [X]" is better than "[X], photo". You likely have to re-game the system a lot to get the result you want. When it does nail things, it is awesome and hilarious though. I asked my trusted friends to try a few prompts... 

Fat drunk guy screaming at kids in a park 

 

 

Fat drunk guy sleeping on a bed made out of beer bottles on the street

 

A painting of the grim reaper stalking his next victim in the night, digital art

Homer Simpson as the grim reaper, scary black and white photo

Edited August 6, 2022 by chim

[Wunderbar]

why does it struggle with faces so much compared to other parts of a human?