Anyone using AI art?

[moofi]

Yes, different method, then to simply find a fitting seed and develop that one usually takes several renderings as well, increasing the overall rendertime for a single picture from minutes to hours. Having one developed image who´s prompt is potentially suited for that one seed, yet might give very different results when used on another seed, is fine if you are satisfied with possibly a few images max a day. Then e.g. I saved around 900 zombie dogs during one day (generated quite a few more) and refined the selection down to 50+ creme of the crops images while around 200 are just a little less good to still pretty good looking and also most of the rest is still fine to acceptable. It´s quite a lot of fun looking at a greater number and variety of zombie dogs anyway, even if the majority of them are not of printworthy quality
Also the potential outcome of hundreds of generations in terms of finding gems is especially interesting. Like finding one seed within something like 400 generations that is very special compared to the others. As well regarding learning how to prompt with that specific AI a greater number of generations makes understanding/exploring easier. The development of an idea is something I´m doing via refining the prompt itself, at least narrowing down the range in that the generation most likely has a favourable outcome. Afterwards I can simply regenerate the identical prompt for basically an arbitrary number of times, delivering varying satisfying images all over the place.
Naturally combining both approaches is favourable after all and something I´m looking forward to. Saying both ways of working have their advantages and disadvantages that can be maximised/minimised when being united.
Anyway, would be interested in your seeding refinement results if you feel like sharing

And yes, rendertimes were a little bit different back then

Yeah, well, currently I´m creating up to 800 images a day, saying 12 minutes rendertime would be a little slow here
With an RTX 3080 you can get one image in around 20 seconds. Even then, though quite acceptable, with wombo I can create a preview grid of 4 pictures within 13-20 secs and only choose the ones I like to render big while the upscaling from a chosen one takes like 1-2 seconds. Then wombo isn´t aiming for seeding, saying it will always be a new outcome generated from the prompt. The seeding plus using varying sampler models is something I´m going for locally eventually when I can afford a new videocard.

Running locally indeed gives more options. As I can use the same seed to re-render the image with little tweaks, generating hundreds of images a day is no real use to me. I like the option to use this to actually develop an idea instead of just randomly generate stuff. For what I (try to) do with it, 12 minutes for a render is acceptable. I used to run DKB-trace on MS-DOS back then (yep, I'm old) and rendering a 320x200 image (the ones with all the shiny spheres) took about 72 hours

[bossa]

No, I don´t.jpg

Very funny!

[Re8et]

It looks like some Apex twin stuff
Get with DiffusionBee on M1 mini 8gb
Interesting stuff, Id like to add my own models but I don't know how.
Sure is it has its own interpretation of words...

piano player walking on a tight rope between two skyscarpers.png (427.72 KiB) Viewed 9283 times

[crimsonwarlock]

Trying to get some nice visuals for my synthwave project.

[Re8et]

Trying to get some nice visuals for my synthwave project.

Way cooler than any of my attempts so far...

[crimsonwarlock]

Trying to get some nice visuals for my synthwave project.

Way cooler than any of my attempts so far...

That one was just a simple prompt: "a fast car at night, synthwave". When using a reference picture you get much better results:





I'm plenty happy with these results

[moofi]

Could be a Pink Floyd cover

It looks like some Apex twin stuff
Get with DiffusionBee on M1 mini 8gb
Interesting stuff, Id like to add my own models but I don't know how.
Sure is it has its own interpretation of words...

piano player walking on a tight rope between two skyscarpers.png

[bxbrkrz]

You don't even have to type anything, just some retouching for some of the visual 'misbehaving'.

[moofi]

In addition, reference images can spice or open up things pretty well, especially when the AI is a little stuck in its standard patterns with a certain prompt. What influence setting did you use with that one?

As well, like mentioned before, adding artists to the prompt is one of the greatest helps. It does not only introduce a cohesiveness the pure prompt is often missing but interpretates the prompt in new/improved ways. Like when mixing objects/beings for example, the merging process works much better with an artist added from what I´ve experience so far.

Some artists to check out besides the obviously very famous ones like da Vinci, Dalí, Picasso, etc.:

CG: Simon Stalenhag

Grandmasters/painters: Jean-Honoré Fragonard, Andrea Mantegna, Pieter Bruegel, Otto Dix

Illustrators: Greg Rutkowski, Gustave Dore, Greg Staples, WLOP, Anton Nazarenko

Photographers: William Eggelston, Gregory Crewdson, Kirsty Mitchell

Various Artists: Edward Gorey, Beksiński


You can as well check out movie titles in there for a certain style. Simply type "title movie"

These ones containing "tron movie":

tron movie.jpg (604.5 KiB) Viewed 9267 times

Way cooler than any of my attempts so far...

That one was just a simple prompt: "a fast car at night, synthwave". When using a reference picture you get much better results:





I'm plenty happy with these results

[crimsonwarlock]

What influence setting did you use with that one?

The same simple prompt, but I added "colorful artstation". That and using a reference picure. I'm using "Stable Diffusion UI", which gives you a nice selection of tags to use for tuning. When you have a useful image, you can stop the seed from randomizing and get the same image while you experiment with the prompt.

https://github.com/cmdr2/stable-diffusion-ui

[moofi]

I ment, what influence setting did you use on the input image?

Thank you for the link btw

What influence setting did you use with that one?

The same simple prompt, but I added "colorful artstation". That and using a reference picure. I'm using "Stable Diffusion UI", which gives you a nice selection of tags to use for tuning. When you have a useful image, you can stop the seed from randomizing and get the same image while you experiment with the prompt.

https://github.com/cmdr2/stable-diffusion-ui

[crimsonwarlock]

I ment, what influence setting did you use on the input image?

There are no setting for the input image in Stable Diffusion UI, only a slider that sets a value for “follow prompt”, which was set at 0.7 or 0.8 (goes from 0 to 1). It seems to balance between the text prompt and the input image, when I set it to a low value I basically get almost the input image returned.

[littlejam]

hello,

this is getting to be an interesting thread
my only problem is that the stable diffusion v2
appears to require windows 10
i run windows 7 pro

is there any way of getting this to run on windows 7 pro?

or, is there another ai imaging program that will run on windows 7 pro?

i have abstract art and real photographs i've taken
that i would like to use as the initial image and go from there
and use the images as tv pilot episode posters for the scripts i'm writing

any advice is appreciated

cheers,

j

[Rising Night Wave]

ok. i downloaded that Stable Diffusion UI v2.1 program.
after many tries and several hours exploring prompts and tags i came up with this image:



very nice. like a pro

here are specs for this image:
ancient city is surrounded by space ships which are firing flames at demons in the apocalyptic sky, Highly Detailed, Very Intricate, Colorful, Concept Art, Digital Art, Fantasy, Cinematic, Dramatic, Film Grain, Landscape
Width: 1024
Height: 512
Seed: 7615869
Steps: 50
Guidance Scale: 7.5
Prompt Strength: 0.8
Use Face Correction: None
Use Upscaling: RealESRGAN_x4plus


--- --- --- --- --- ---


am, only one thing is bothering me:
FFS!!!



hesus christus eating all macaroni!!! are they serious?!!


--- --- --- --- --- ---


otherwise... this is gonna be an interesting journey.
this app is sooooo coooool.

thanks for mentioning this app.

[moofi]

Well, wouldn´t have thought so either, then it´s a local AI being fed with something like 5.8 billion images with 800 million parameters if I remember correctly. It merely having 800 million parameters "only" is the reason for it to be able to run on a local standard PC in the first place.

ok. i downloaded that Stable Diffusion UI v2.1 program.
after many tries and several hours exploring prompts and tags i came up with this image:



very nice. like a pro

here are specs for this image:
ancient city is surrounded by space ships which are firing flames at demons in the apocalyptic sky, Highly Detailed, Very Intricate, Colorful, Concept Art, Digital Art, Fantasy, Cinematic, Dramatic, Film Grain, Landscape
Width: 1024
Height: 512
Seed: 7615869
Steps: 50
Guidance Scale: 7.5
Prompt Strength: 0.8
Use Face Correction: None
Use Upscaling: RealESRGAN_x4plus


--- --- --- --- --- ---


am, only one thing is bothering me:
FFS!!!



hesus christus eating all macaroni!!! are they serious?!!


--- --- --- --- --- ---


otherwise... this is gonna be an interesting journey.
this app is sooooo coooool.

thanks for mentioning this app.

[moofi]

Pleasant picture

ok. i downloaded that Stable Diffusion UI v2.1 program.
after many tries and several hours exploring prompts and tags i came up with this image:



[...]

[moofi]

Well, wouldn´t have thought so either, then it´s a local AI being fed with something like 5.8 billion images with 800 million parameters if I remember correctly. It merely having 800 million parameters "only" is the reason for it to be able to run on a local standard PC in the first place.

hesus christus eating all macaroni!!! are they serious?!!


--- --- --- --- --- ---


otherwise... this is gonna be an interesting journey.
this app is sooooo coooool.

thanks for mentioning this app.

[/quote]

[bxbrkrz]

Yep. 5.8B

https://haveibeentrained.com/

[moofi]

No, I don´t.jpg

Very funny!

[littlejam]

hello,

i run windows 7
i asked earlier in this thread about running ai imaging programs on windows 7
nobody responded

now, i am asking if somebody can make me an image
if i give them words / phrases
based on the plot of the movie episode

thank you in advance

should be a superdope image

cheers and eat well,


j

[bxbrkrz]

[rsnsnds]

Would you consider AI music your music if the only thing you would have to do would be to push a button ?

[platzangst]

Would you consider AI music your music if the only thing you would have to do would be to push a button ?

No, I'd consider that the radio.

[rsnsnds]

Would you consider AI music your music if the only thing you would have to do would be to push a button ?

No, I'd consider that the radio.

I was talking about making music, not listening to music.

People don't think that the music on the radio is their music, like people pushing a button to create AI art


But you're actually right, AI art is like radio for art, as you are only admiring the result, without actually doing any work

[platzangst]

People don't think that the music on the radio is their music, like people pushing a button to create AI art


But you're actually right, AI art is like radio for art, as you are only admiring the result, without actually doing any work

There are a couple things to get into here, though:

It's not like the people using AI for art simply "press a button" - in most cases there's some sort of prompt, a concept that the AI attempts to realize. You have to give the AI something to work with. In that sense, a person could be said to be doing some work, even if it's all on the conceptual side, without any of the technical skills.

In the same sense, if I were to be able to enter some text prompts into an AI and get music, that would be a bit more than "pressing a button". If I said, "give me a smooth yacht rock tune blended with greasy glitch IDM in the key of E at 95 BPM" then I'd be providing an idea to work from, instead of passively activating a random music machine.