Star Trek AI art thread

[publiusr]

Text to video now?
https://www.nextbigfuture.com/2023/03/gen-2-ai-text-to-video.html
https://www.nextbigfuture.com/2023/...ike-ui-to-your-downloaded-language-model.html

AI audio
https://www.nextbigfuture.com/2023/05/artificial-intelligence-for-music-production.html
https://techxplore.com/news/2023-08-humans-unable-quarter-deepfake-speech.html
https://phys.org/news/2023-08-speech-deepfakes-frequently-people.html

Help for disabled artists
https://www.nbcnews.com/tech/tech-news/brain-waves-ai-can-sketch-picturing-rcna76096

Zelig on steroids
https://www.nytimes.com/2023/04/08/business/media/ai-generated-images.html

Hugging Face
https://www.nextbigfuture.com/2023/...openai-will-offer-chatgpt-business.html?amp=1

Coming
https://www.nextbigfuture.com/2023/...de-building-to-superhuman-level-ai.html?amp=1

More:
https://phys.org/news/2023-08-people-emotions-computer-generated-art.html
https://techxplore.com/news/2023-07-ai-image.html
https://techxplore.com/news/2023-07-3d-display-digital-world.html

 

[Captain pl1ngpl0ng]

Text to video now?

yea i tested some a good while back (a week or so in AI language is a good while back)
anyway, it's very far from good, not even ok... it takes a looong time (40 sec)
to generate a 2 sec clip, and the result is garbage.
but yea give it a few weeks, its coming..... making gifs for memes and avatars going to be funny.
meta are also involve like everybody else: https://makeavideo.studio/

this is stable diffusion:
edit: https://huggingface.co/spaces/damo-vilab/modelscope-text-to-video-synthesis
i tried : "dog riding a duck"

 

[somebuddyX]

OOH yes more of this please!

 

[Cyfa]

Characters from the Disney animated film "Starships"

~ Adorable!!!

[I'm on the fence about AI generated images. I think they're a great starting off point for a human/person to produce a piece of art based on the generated image. And any AI generated images for anything other than personal use should include a list of artworks (and their artists) that were used to form it.

I read this article about how AI art is generated (amongst other things) in the Guardian the other day which I found rather informative]

 

[Captain pl1ngpl0ng]

[I'm on the fence about AI generated images. I think they're a great starting off point for a human/person to produce a piece of art based on the generated image. And any AI generated images for anything other than personal use should include a list of artworks (and their artists) that were used to form it.
I read this article about how AI art is generated (amongst other things) in the Guardian the other day which I found rather informative]

I didnt read the article, but i wholeheartedly agree with listing other art/artists name if this is used in the instructions/prompts to generate an image.
However it is not at all important to use such names in the instructions to produce great images, certain triggerwords or even the order you choose to write something in can make a huge different in the outcome.

I also understand that (real) artists are sceptical about this new thing. I guess that this will eventually be a natural part of a workflow. As a tool to create concepts and random ideas, the AI are actually fantastic. But it stops there....
An art director cant work with a team of non-artists that are "good" at AI, and only AI. A team with good artists AND the knowledge of AI will certainly be the natural evolution here.

AI is really fun as creative outlet for folks like me with amputated artistic skill sets, and also for more advanced artists...
and it will be an important tool for professionals, that's my take on this whole thing anyway.
Come over to the dark side @Cyfa

 

[somebuddyX]

~ Adorable!!!

[I'm on the fence about AI generated images. I think they're a great starting off point for a human/person to produce a piece of art based on the generated image. And any AI generated images for anything other than personal use should include a list of artworks (and their artists) that were used to form it.

I read this article about how AI art is generated (amongst other things) in the Guardian the other day which I found rather informative]

My only "issue" and it's not really one but when I do stuff in 3D I can spend time with the same character or look at things from different angles and at the moment I don't think I get that with AI art. But that's just my way of working. I make more like several page "comic" stories than like one single poster. One of the things I do love about AI art is that it's really fun to use as background art in a scene. When I do scenes in front of a cinema or posters in someone's house I used to spend time making dumb fake posters in GIMP that were only in the background and would take up time that I could have spent working on the actual scene. Now I can make a poster with a few prompts, throw on some text and then the whole thing's done in a few minutes. For me it's no different than how people would make mashups in Photoshop or Gimp but just more instantaneous.

 

[blssdwlf]

~ Adorable!!!

[I'm on the fence about AI generated images. I think they're a great starting off point for a human/person to produce a piece of art based on the generated image. And any AI generated images for anything other than personal use should include a list of artworks (and their artists) that were used to form it.

I read this article about how AI art is generated (amongst other things) in the Guardian the other day which I found rather informative]

After reading the article I think one of the things the author lightly touches on is that the AI "hallucinations" and twisted imagery are from the datasets were lifted from the whole of the internet which suggests that there probably is some part of the internet filled with outright lies and twisted imagery. Supposedly contractors were hired to filter out the crazy stuff but I seriously doubt they would've made a dent in it.

I would love to see some way to debug an AI's responses to know what images and text it used as sources.

 

[Cyfa]

I didnt read the article, but i wholeheartedly agree with listing other art/artists name if this is used in the instructions/prompts to generate an image.
However it is not at all important to use such names in the instructions to produce great images, certain triggerwords or even the order you choose to write something in can make a huge different in the outcome.

~ Sorry, I didn't make myself clear in my previous post. I meant that the listing should be of artworks (& artists) that the AI "lifted from the whole of the internet" (as @blssdwlf mentioned) in order to generate the final image. But yes, you bring up a good point that I didn't think of in that certainly if any of the prompts were "make a picture in the style of Salvador Dali" for example, then the final piece should acknowledge that
I totally agree with your other points such as "a tool to create concepts and random ideas" and "creative outlet for folks like me with amputated artistic skill sets" (I'm not sure if you mean for amputees/differently abled people, or that it's the artistic skill itself that's just not there?* But either way)

* I've seen your work, and there's definitely artistic skill!

After reading the article I think one of the things the author lightly touches on is that the AI "hallucinations" and twisted imagery are from the datasets were lifted from the whole of the internet which suggests that there probably is some part of the internet filled with outright lies and twisted imagery. Supposedly contractors were hired to filter out the crazy stuff but I seriously doubt they would've made a dent in it.

I would love to see some way to debug an AI's responses to know what images and text it used as sources.

~ Yes! That's what I meant: what images from the internet did the AI use to create its final image. Thank you!

One of the things I do love about AI art is that it's really fun to use as background art in a scene. When I do scenes in front of a cinema or posters in someone's house I used to spend time making dumb fake posters in GIMP that were only in the background and would take up time that I could have spent working on the actual scene. Now I can make a poster with a few prompts, throw on some text and then the whole thing's done in a few minutes. For me it's no different than how people would make mashups in Photoshop or Gimp but just more instantaneous.

~ Oh, yes, I agree and think that backgrounds and set dressing is a great use for AI generated images. What a time saver (as you pointed out)!


I'm now wondering if the AI can be instructed to use explicitly free/open source/public domain images only to generate its images?

 

[Captain pl1ngpl0ng]

I meant that the listing should be of artworks (& artists) that the AI "lifted from the whole of the internet" (as @blssdwlf mentioned) in order to generate the final image

ahh.... yes, that would be great. I would love to see what was behind the images produced.

folks like me with amputated artistic skill sets" (I'm not sure if you mean for amputees/differently abled people, or that it's the artistic skill itself that's just not there?* But either way)
* I've seen your work, and there's definitely artistic skill!

i meant that the skill set is not there yes, and also thank you for the compliment
i have called you a true artist in the past because you actually can draw... i can't=)
but i guess i have an artistic sense at least, so thank you.

I'm now wondering if the AI can be instructed to use explicitly free/open source/public domain images only to generate its images?

i dont think that is possible right now from what is available ...
but, i know that very soon Adobe are coming with their AI-soulution,
and their AI are trained only on Adobe Stock Images which are royalty free stuff from adobe users,
and i have read somewhere that also other tools are being trained with public domain stuff.
--

when I do stuff in 3D I can spend time with the same character or look at things from different angles and at the moment I don't think I get that with AI art

i had the same feeling at first, "write some words, click a button, hope something good comes out"....
that didnt really gave me anything. But then i realised i could be much more involved in the process.
AI generally spits out crazy stuff with lots of flaws like 12 fingers on one hand and merged bodies,
or something thats kind of right but maybe stuff like composition etc is missing,
so i thought if one look at stuff from a curators mindset and recognize what have a potential and are able to
work with the image with other tools like photoshop and then go through AI again, back and forth like that,
then i at least got the illusion of being a part of the final result.

Example: im working on an abstract series now with swimmers, the picture to the left are straight out of AI,
the picture on the right are dragged several times through the inpaint part of AI, then into photoshop to adjust stuff manually, (cut and paste and move stuff etc) because AI dont understand everything,
and then back again into AI for final renders.

So even tho i didnt do anything that i would call artistic,
i was definitely a large part in pushing this in the right direction, and i can fool myself saying "I did that"

 

[somebuddyX]

So even tho i didnt do anything that i would call artistic,
i was definitely a large part in pushing this in the right direction, and i can fool myself saying "I did that"

I definitely think this stuff is artistic! And I love the idea of merging stuff in Photoshop. I was trying that with dream by building a futuristic city by cutting it up into pieces and putting it through and then reassembling it in GIMP but the lighting of it didn't merge well.
I just mean for me sometimes I will be in a scene in Poser and then switch to a different camera and get a different view on the same scene and then realise it works much better. Or I get attached to a character I made in dream and then realise I'll never see them again. I think that will change in time with more controls and AI learning. People make Star Trek scenes and there's a randomness to the uniforms and commbadges especially but I think over time with maybe AI learning what the creator wants we'll see people be able to make things more consistent. I think of Tony Stark in "Iron Man" with that interface building Mark II suit and that's what I'd love to see in the future. But instead me and Jarvis making Star Trek TNG Season 8 episodes.

 

[publiusr]

I am hearing about LeiaPix…which turns 2D into 3D…
https://convert.leiapix.com/

Here are other tools coming out:
https://www.nextbigfuture.com/2023/04/over-1300-new-ai-tools-including-codeium-and-leiapix.html

Solar system humanoids
https://curiosmos.com/mind-blowing-...inhabitants-of-our-solar-system-created-by-ai

 

[Captain pl1ngpl0ng]

also keep an eye out for this one, txt to vid, txt+img to vid, vid to vid.
https://research.runwayml.com/gen2

 

[publiusr]

Now you can pull images out with this tool:
https://segment-anything.com/
https://techxplore.com/news/2023-04-powerful-meta-ai-tool-individual.html

This is great for doodles
https://techxplore.com/news/2023-04-generative-tool-2d-3d.html

Now for AI to makes these covers live
https://www.perrypedia.de/wiki/Titelbildgalerie_PR_1_-_99

Image extractor
https://techxplore.com/news/2023-04-power-efficient-disentangle-visual-attributes.html

 

[somebuddyX]

Maybe this has been posted but Civitai has some Star Trek Lora including uniforms
I haven't tried them but I might give the uniforms a go.
https://civitai.com/tag/star trek

 

[Captain pl1ngpl0ng]

^ yea i havnt seen these, especially the insignias looks very consistent

 

[publiusr]

For human fashions, this find may be of use:
https://techxplore.com/news/2023-04-human-prompts-generate-realistic-fashion.html
In initial evaluations, the model created by this team of researchers achieved very promising results, creating realistic images of garments on human bodies inspired by human sketches and specific text prompts. Their model's source code and the multimodal annotations they added to the datasets will soon be released on GitHub.

DANSE CYBRE
https://techxplore.com/news/2023-04-ai-powered-animator-generative-ai-choreography.html
Stanford University researchers have developed a generative AI model that can choreograph human dance animation to match any piece of music. It's called Editable Dance GEneration (EDGE).

 

[somebuddyX]

 

[somebuddyX]

I put "van gogh starry night style painting of star wars"

 

[fireproof78]

Interesting. What were the prompts? I'm still learning this AI generation thing, since my art skills are terrible, but my prompts never go far.

 

[somebuddyX]

Interesting. What were the prompts? I'm still learning this AI generation thing, since my art skills are terrible, but my prompts never go far.

I went here and downloaded and used the LORA and I think I just copied one of the prompts from one of the existing pictures and went through a bunch of different models trying things. I got really lucky with the first picture. Civitai has a bunch of cool checkpoint models and I'll just go through and see what's top rated and get them, or I set it to just show images and see what other people have made and then try and find the models and use their prompts. Realistic Vision was the first model I used and I really like it.
The thing with Trek stuff and genre stuff in general is that it's so up in the air whether it'll work or not and it might take a bunch of images to get one decent. But you can find some fun LORA and prompts to make some cool stuff at civitai. I saw one that's like images in a kid's book that I want to try.