I was just messing around with an art generator at Night Cafe. Seems pretty simple.
Has anyone done any Traveller art with the AI tools out there?
Has anyone done any Traveller art with the AI tools out there?
General Has anyone done any Traveller art with the AI tools out there?
- Thread starter Spinward Scout
- Start date
kilemall
SOC-14 5K
Left leg looks amputated to me.
I'll have to try that. Does it cost anything?
FreeTrav
SOC-14 1K
No, but you do have to sign up. You accumulate 5 credits per day free; that pic only cost 1, but it's possible to cost more than that. You can also buy packages of credits for real money.
Timerover51
SOC-14 5K
Has anyone tried Google Gemini?
Haven't heard of that one.
FreeTrav
SOC-14 1K
I think at the moment Google has taken it off-line; there was a furor about it being asked for images of German soldiers during the late 1930s, and it made them all non-white.
It got a bit deeper than that - pretty much seems (at least from reporting) to have a built in "diversity filter" for want of a better word - searches for images of Vikings returning all non-whites, Popes being male & female non-whites, and so forth. Not sure how accurate, but some outlets were saying it was almost impossible to get Caucasian images for historical figures.
Timerover51
SOC-14 5K
Those comments are all quite accurate, as they are supported by the posted images.
A Medieval Knight and a Catholic Pope.
Last edited:
Condottiere
SOC-14 5K
This is likely moving towards a paid model, since at some point it's going to get monetized.
At which point, the would be artist determines parameters.
At which point, the would be artist determines parameters.
FreeTrav
SOC-14 1K
I subsequently heard about that, but at the time I posted, I'd only seen about the multiethnic Nazi troops...
Kakistocrat
SOC-12
To be fair, there are currently at least 16 elector cardinals of sub-Saharan African descent (including a Haitian, excluding a Moroccan and a South African), so one of them being elected pope in the future is conceivably within the realm of possibility. (I think that there have been three Catholic popes from [Roman] Africa, but the most recent of them was around AD 500, and all of them were likely of at least partial Berber descent.)
Condottiere
SOC-14 5K
Science fictionwise, less of an issue.
The phenomenon not only effects Google, but also the current crop of historical (docu/melo)dramas.
Even when you play it safe, you end up saying sorry.
The phenomenon not only effects Google, but also the current crop of historical (docu/melo)dramas.
Even when you play it safe, you end up saying sorry.
But it's not as though it hasn't gone the other way for centuries as well, in cinema as well as other media -- with much higher stakes and influence.
This guy, for example:
(Image credit: wikipedia)
Keeping in mind the demographics of the region, and his culture of origin -- is it at all likely he looked like that?
No AI involved, just cultural adaptation.
Last edited:
But it is an issue -- almost entirely because these systems have no concept of what it is they are generating.
It's why you get extra fingers a lot of the time: what is most likely to be next to a finger when representing a hand (and it does not know what a hand is, just that it's an element likely to appear in certain areas of pictures of people) is another finger.*
It does not know what those troops are, or why that would make them intrinsically non-diverse.
We see the constraints that the system does not, and cannot understand.
----------
*Have you ever really looked at your hand, maaaaan.... (I'm kidding, but...) Looking at a hand, 50% of fingers are next to other fingers, while the other 50% are next to whatever is beyond the hand. So, place the first finger in the image, and start rolling dice to place more fingers until you come up with a finger that's "next to background" and stop there. It's a vast over-simplificatuon of how it really works, but it's a good analogy nonetheless.
It's why you get extra fingers a lot of the time: what is most likely to be next to a finger when representing a hand (and it does not know what a hand is, just that it's an element likely to appear in certain areas of pictures of people) is another finger.*
It does not know what those troops are, or why that would make them intrinsically non-diverse.
We see the constraints that the system does not, and cannot understand.
----------
*Have you ever really looked at your hand, maaaaan.... (I'm kidding, but...) Looking at a hand, 50% of fingers are next to other fingers, while the other 50% are next to whatever is beyond the hand. So, place the first finger in the image, and start rolling dice to place more fingers until you come up with a finger that's "next to background" and stop there. It's a vast over-simplificatuon of how it really works, but it's a good analogy nonetheless.
Last edited:
Commander Truestar
SOC-13
it's interesting seeing the results people have gotten...
....but, what prompt was used to get the results?
....but, what prompt was used to get the results?
I've found with my dabbling, but it's been awhile, that I don't have the patience to argue with the AI to get the prompt just so, and, in the process, consume however many free credits they're willing to give me. I've found the image space quite frustrating to work with, so I've stopped playing with it myself. That said, the examples posted here are quite nice.
At this stage, though, I prefer the stuff from the (assumedly) Carbon Base Lifeform @magmagmag .
At this stage, though, I prefer the stuff from the (assumedly) Carbon Base Lifeform @magmagmag .
