I don’t have the latest version of PAI so for fun I ran your result through Luminar NEO with a 2X upscale and then Super sharp with face enhance. I did it once and here is the result.
1 Like
But the problem is that my restoration was 4 times the resolution you upscaled. I don’t know why the photo was reduced in resolution by 4 when I imported it.
Next time, I’ll use an image link because I have the impression that the site has a size limit for image resolution when importing.
Ici, ma mise à l’échelle est encore meilleure. J’aurais dû réduire la résolution de 4 et l’augmenter de 6 par la suite.
Now everything is perfectly sharp without too much exaggeration.
True about the PAI painting look, but my last example was further artsified with Exposure’s Snap Art 4.
1 Like
I Just Downloaded The I Mac Version.
I too am getting the freezing on face detection. I have 2 computers. This one is a newish computer, but with no real GPU. So it uses the computer’s power to work. So naturally it is slow. Which I fine. Some photos it works fine with face detection. Others it fails all the time.
I am running 1.3.6 on my machine machine which has a proper GPU with 6gb of ram and is quick. I keep the old version on there because I find it works better for most of the photo. Mainly because anything newer tends to detect with lens blur model and the sharpening with that model makes my photos look like crap. The standard or strong model look better. But with 2.3.2 when if it decides to use lens blur and I manually override to use standard, I get these squares around the subject in the background. I have subject detection turned off. Also as mentioned in an earlier post, in some photos it randomly changes the color of a hat. The hat is all black, but after PAI runs, it shows up partially gray in weird patterns
Thx for your kind note!
I debated, but was nervous crawling around on the ground right under where the tree fell and causing any vibrations trying to heft the desktop PC tower, monitor, accessories out of there… when the FD said it was the most vulnerable room to more damage I stayed out. And, placed a Buddha just outside the door in the hall to keep the room safe.
An arborist came midday the day after the storm/tree fall. They got it off the house and chopped most of it down. They left the big stump & root system and the boulders embedded in it.
I was thinking I should use Ps & PAI to create a line drawing I can ‘tattoo’ onto the top of the sliced off stump.
Roofers were here yesterday. Now waiting for quotes for repairs. Lots of expenses (in addition to the trauma) when we’re not working.
Yeah, that’s just right!
I have the same problem!
So we’re, now at more than a week since I first reported this issue, and several others have reported it as well.
Has there been any progress toward fixing it?
Attn: @anthony.lawn
1 Like
Happily, it’s still working for some of us. My prayer is that they won’t fix it by breaking it for me.
1 Like
Enlarging and cleaning up a DiffusionBee AI render in Photo AI.
Original render:
PAI cleaning up the lo-res:
The next task was to improve the eyes. Here is the first pass in PAI:
Better, but still not there. I performed additional attempts in both PAI and Photoshop but both got worse from there ;-). This is the best of the alternates in PS:
This was the keeper (PS):
After further trips through Photoshop (generative fill), Photo RAW, Radiant Photo and back into PAI for final cleanup, I finished with this!
Next up is a real iPhone 12 photo shot yesterday while running errands in SoCal (we keep our snow at a distance…).
PAI was used to remove distractions and enlarge/smooth the original (the capture was zoomed in on the phone which is never a good thing…).
Additional treatment with the aforementioned apps plus Snap Art 4 made the result “mostly” usable:
Very nice results but it seems that it is only for Macs. I am a paid user of Leonardo AI and have created many artworks.
2 Likes
That’s really nice! I love banging out AI landscapes 
Here’s an un-reprocessed example from DiffusionBee:
I see Leonardo AI is web-based. I enjoy the local processing of my renders (M2 Mac Mini with various downloaded models, takes less than a minute per). I have also tried some online options; there are so many choices now to get this stuff done!
Of course you need PAI or GAI to upsize and smooth these renders as 1024 seems to be a limit (at least in DB).
1 Like
I do use Topaz PAI to upscale. One issue I found is that if an image contains people then sometimes the AI will put too many fingers on hands and distort feet and hands. Not every time but this can ruin an otherwise great picture. Some can be corrected or covered but not always.
Here is one example with people but where that hands and feet are ok.
Yeah, but it’s not so bad.
I love this rendering. And the colors really pop.
Another beautiful work! And yes, follow-up touchup is usually needed… Which is good, it adds some human element to the piece.
Besides the eyes, my blues girl had weird fingers and the guitar strings were like wet spaghetti! Plus the bottom of the image was truncated in the original render and needed expansion. The original mood lighting was fine but I chose to glam it up for the clubgoers
Do you have the original prompt and seed for your generated image? I’d like to see if I can do better in Fooocus AI.
