Post your AI images
Discussion
I've given the CoPilot part of the bing app a go with ChatGPT. First prompt was 'A Stormtrooper Smurf'
Then added 'make it carry a blaster' to the chat, as the 4 originals had lightsabers. I picked out this one as the helmets went a bit wrong with the others.
Then added a style to it
It's got filters too to blur backgrounds or make parts black and white. Interesting, and a bit of a different way of working than the Bing Creator.
Then added 'make it carry a blaster' to the chat, as the 4 originals had lightsabers. I picked out this one as the helmets went a bit wrong with the others.
Then added a style to it
It's got filters too to blur backgrounds or make parts black and white. Interesting, and a bit of a different way of working than the Bing Creator.
WookieNutz on Facebook posted this phooshop, I tried doing a version of it with Copilot.
The prompts for this went...
princess Leia in a yellow jumpsuit carrying a blue light saber, on a bright yellow background
make it a full length body photo
Make her pose confidently
Then a few prompts to try sorting the text...
add a title in black text "KILL JABBA"
make it a movie poster
one more prompt for the picture
Change her expression to be more intense
Made the bottom right picture of the last set landscape, and got this
Close enough with the text on kill, I'll take it.
The prompts for this went...
princess Leia in a yellow jumpsuit carrying a blue light saber, on a bright yellow background
make it a full length body photo
Make her pose confidently
Then a few prompts to try sorting the text...
add a title in black text "KILL JABBA"
make it a movie poster
one more prompt for the picture
Change her expression to be more intense
Made the bottom right picture of the last set landscape, and got this
Close enough with the text on kill, I'll take it.
Nice to see it ignores the most basic of request details over half the time 
When you can add details like full length/full body and actually get it from all outputs, I’ll start to take it a bit more seriously.
For now it’s like working with a super talented dreamer art worker… results are well drawn, but often not what you actually asked for.
When you can add details like full length/full body and actually get it from all outputs, I’ll start to take it a bit more seriously.
For now it’s like working with a super talented dreamer art worker… results are well drawn, but often not what you actually asked for.
Mr Whippy said:
… results are well drawn, but often not what you actually asked for.
These two camera girls are a case in point. Both use exactly the same prompt.Bing version gave me more or less exactly what was asked for:
Imagebot version totally forgot about the photo and all the other photos round it, also it got every colour wrong
:
I have to say though, that the Imagebot looks more realistic. Also, quite often Bing will miss out either the camera or glasses - frequently both.
They really need to give us more assistance in how the prompts should be worded. Sometimes making great changes in the text achieves nothing, others just a one or two word difference can wreak havoc.
Stealthracer said:
Mr Whippy said:
… results are well drawn, but often not what you actually asked for.
These two camera girls are a case in point. Both use exactly the same prompt.Bing version gave me more or less exactly what was asked for:
Imagebot version totally forgot about the photo and all the other photos round it, also it got every colour wrong
:
I have to say though, that the Imagebot looks more realistic. Also, quite often Bing will miss out either the camera or glasses - frequently both.
They really need to give us more assistance in how the prompts should be worded. Sometimes making great changes in the text achieves nothing, others just a one or two word difference can wreak havoc.
This is the issue I noted previously though.
It’s based on LLM and arbitrary metadata tags on platforms like behance, artstation, instagram or wherever they’re scraping… no true consistency.
ajprice said:
The prompts for this went...
princess Leia in a yellow jumpsuit carrying a blue light saber, on a bright yellow background
make it a full length body photo
Make her pose confidently
Then a few prompts to try sorting the text...
add a title in black text "KILL JABBA"
make it a movie poster
one more prompt for the picture
Change her expression to be more intense
princess Leia in a yellow jumpsuit carrying a blue light saber, on a bright yellow background
make it a full length body photo
Make her pose confidently
Then a few prompts to try sorting the text...
add a title in black text "KILL JABBA"
make it a movie poster
one more prompt for the picture
Change her expression to be more intense
Edited by Cotty on Monday 15th April 09:05
ajprice said:
Stealthracer said:
I just tried Copilot but couldn't even figure out how to get an image to appear,
Don't do the whole descriptive detailed prompt thing, give it a simple description of something.Just give it a simple description and be happy with what you get given.
This is the AI revolution
me!Impressive, but ultimately useless.
For now it’s a spectacle, but I’m still struggling to get it to adhere to very clear explicit prompts like ‘white background’
I’d be more impressed if it could come up with an entirely new range of IP that passes the ‘not just copied off art station’ sniff test, and then make up a novel story, and then make an advert for that film and be proven to show that direction by a human director was followed.
Otherwise what use is it?
I just paid £10 for a month of getimg.ai to give it a go on a project.
I think the ideas themselves are good, but then it’s like a brick wall.
Tidying up work, getting stuff like colours correct, etc, is wasting hours.
I’m literally using it for ideas then redrawing at this rate.
To sum up, looks impressive. But is as shallow as a puddle for now.
For now it’s a spectacle, but I’m still struggling to get it to adhere to very clear explicit prompts like ‘white background’
I’d be more impressed if it could come up with an entirely new range of IP that passes the ‘not just copied off art station’ sniff test, and then make up a novel story, and then make an advert for that film and be proven to show that direction by a human director was followed.
Otherwise what use is it?
I just paid £10 for a month of getimg.ai to give it a go on a project.
I think the ideas themselves are good, but then it’s like a brick wall.
Tidying up work, getting stuff like colours correct, etc, is wasting hours.
I’m literally using it for ideas then redrawing at this rate.
To sum up, looks impressive. But is as shallow as a puddle for now.
Gassing Station | Computers, Gadgets & Stuff | Top of Page | What's New | My Stuff