Upscaling AGI screenshots

[notbobsmith]

Yeah it is remarkable how much it got right, though. I couldn't paint that screen into a logical painting like it did.

I did laugh on the PQ - where it turned Sonny into a woman in white.

True, but then AI has no idea that Sonny was a man. We do. All that AI sees is a figure in white with its back to the camera.

I think it also noticed the white hat and put Sonny in a summer dress that would be appropriate for the hat.

I picked Quest for Glory II since it is SCI0 and only 16 colors and so closer to the AGI examples just with a higher resolution. I thought the plaza in Shapeir would be a good choice since it has text.



And this is the result:



So it misreads the text. "Katta's Tail Inn" becomes "Antrodic Tal Spa". Also gets "Ali Chica" wrong. Interesting that it struck out more with the text than the AGI images. I guess maybe the size of the text is small and less legible? The Katta snake charmer (and I had a feeling it would have problems with non-human characters) turned into a dragon-man with a hookah. At least a hookah is appropriate for the setting.

[Rath Darkblade]

Thanks, NBS. *scans the image*

The Katta on the left also becomes a man. I wasn't sure about the Katta on the right - I thought it was some kind of lizard-man sipping a drink? Depends on what that spirally thing between him and the drink/hookah is.

I'm not sure why they gave Ali Chica (or should I say Al-Chia???) a beard. In the original image, he's not bearded at all.

Lastly, ChatGPT got the hero all wrong. He shouldn't be tall and gangly, his sleeves should not be rolled up, and his shoes should curl to a point -- after all, it's an "Arabian Nights" setting.

(Oh, yeah: I won't even touch "Antrodic Tal Spa". What does "antrodic" mean, anyway?)

[Tawmis]

Using Copilot - without giving it a reference - I told it to redraw Larry Laffer more realistic...

And aside from the extra bird head in the sign... not bad.

[Tawmis]

So then I uploaded this to Copilot and told it to make it realistic - and it did great - just made Larry far more handsome...



into...

[notbobsmith]

The cover image looks pretty good, except for the garbled text and extra seagull head that you noted. The person even looks vaguely familiar.

The city scene also looks pretty good. I'm not familiar with LA's skyline, so does anyone know if that cluster of buildings real? It reminds me a little of GM's Renaissance Center, but that's in Detroit.

[notbobsmith]

I thought this was kind of interesting. From "Gold Rush!":

The original screenshot:


Now the same scene from the 2014 remake:



And what ChatGPT generated from the first image:



Again, I'm impressed that it was able to make sense of the grocery store sign. "Grocery" honestly looks like "GRCCEEY". But both the remake and ChatGPT get the first word wrong. It should be "Fast Grocery". The first word isn't very legible, and the only way to learn this is by typing "look sign". But it is funny that both the AI and the people who recreated the scene read "First".

The building is missing the skylights in the AI version, but I think the buildings overall look better.

I'm not sure why ChatGPT made this image a portrait. Everything is kind of bunched up.

[Rath Darkblade]

I agree, Larry is much too handsome in the LA shot. And what happened to his world-famous Leisure Suit?

The LSL2 "cover" is better, but again, Larry is missing his leisure suit (and, of course, there's the extra gull head). And if I remember, in the original, the sea gull is rolling its eyes.

[Tawmis]

The LSL2 "cover" is better, but again, Larry is missing his leisure suit (and, of course, there's the extra gull head). And if I remember, in the original, the sea gull is rolling its eyes.

In the cover of LSL2, Larry is indeed wearing a floral shirt (very similar to the AI image) and not his Leisure Suit. So it got that right.