we are kidding ourself arnt we

Gigapixel AI didn't do well with the eyes tbh

Another version with a different look by using a high pass filter sharpening and setting the dresses to white in photoshop elements.

pixel for pixel it looks identical

It still looks historical to me but in a different way.

But the dresses are much whiter on my monitor.

As I said originally, as long as you are happy that is all that matters.

I agree. The eyes of the little girl in the front center are messed up.

It's worse than that girl. Top row: ##2 and 6, Middle row: ##1, 4, 6, front row: that little girl. Some of them are minor, but it's not ideal when one eye turns totally black.

Here's the original run through Topaz Photo AI with Autopilot making the decisions, and upsized to four times the pixel count. It maybe looks a little too clean and sterile, but that can be adjusted to personal taste. I have the preferences set to produce 'midrange' results and didn't make any further tweaks one way or the other. Notice that the face recovery feature has placed a tiny highlight reflection in most of the subjects' eyes and added some mild coloring to some of the eyes and lips.

We can be fairly sure that Photo AI has never seen your mother's photo before, so I'm curious to know how close it came to reality with this guess.

It can do some amazing things with faces, but not so much with shoulders and collarbones. Those areas would require some manual corrections.

the image is close just the neck looks a bit off.

Doing that is a step too far for me!

David

Yes, it shows whether more pixels is better or not depends on the contents of the image, the print size and viewing distance.

Viewing that image at arms length as a smallish sized print the flaws in the eyes might not be noticeable but on a much larger print at arms length they probably would be.

More original pixels would most probably improve the quality of a large print.

Which - the added coloring? That was kind of unexpected, but it's subtle and I suppose it can be undone manually. I think the face recovery in general is pretty amazing.

Doing this is quite arbitrary in my opinion.

David

Okay, my opinions:

The catchlights definitely help. There were probably some present in the original photo, but they became lost with the low resolution, artifact-heavy version that was posted.

It would be nice if the software could be told not make those slight color changes, although they can be undone with a bit of manual work. Actually, I've never seen that happen before. Maybe it has something to do with the sepia tint.

I'd say despite a bit of overreaching, the Photo AI results are far superior to anything else posted in the thread so far.

im amazed that there is hair detail and eye lash detail ,did AI put that in ?

That depends on what the aim was and what someone likes and dislikes.

Thank you for your unsolicited but excellent illustration of the "bleeding obvious".

David

If you mean like your unsolicited reply to my post to sybersitizen then you're welcome 🙂

Photo AI did everything you see. I made no changes to its output.