Face recovery is too aggressive in most instances, and frequently detects good quality faces as ‘low quality’. The resulting recovery wipes out most fine details (see attached example). I find myself turning off Autopilot mode and face recovery more often than not.
Face detection and recovery should have to be turned on manually as currently I only see it working decently for small faces for example group shots, crowds etc.
Setting the only filter to ‘enhance - natural’ and resizing by 2x does produce amazing results. Enhance-natural is the best AI filter I’ve seen to date on any Topaz product.
No, I haven’t downloaded them but I guess I will even though the change should be obvious and it isn’t. Even Vargolsoft said to only click on the arrow to change the picture and thats what I didi. My monitor is a calibrated ASUS Proart 1920 x 1080 rez using a Displayport connection. I’ve done photo work for many years.
OK, I copied the two pictues ad loaded theminto Affinity Photo. Enlarging them I could see the cleared faces on three.
Hey ! It has been reported before here, but it’s certainly not bad to tell it more, but certainly the devs are aware of it. in ALL the case where a face is identified, the face recovery is always at 100. As amazing this feature is, each time a face is recognisez, in 99% of the case, like you tell, the 100 is too strong and not adapted. so don’t know what they planned, we’ll see in the next version certainly.
Maybe the face recovery % will be handled better and not implanted fully, or… anyway in 99% of times, i have to switch off the Auto and lower a lot the face recovery setting.
Wow Weird… but i can assure you that the change between the two pictures, even if they are subtle are noticiable, and i have a low (unfortunatly) quality monitor (AOC 24"),
i have the eyes (and glasses) of my age (lol).
I’m watching my monitor through a plexiglass (no kidding ! so it’s kind of i’m watching the screen behind a mirror with a bit lost in quality) and i have the warm color activiated on windows to reduce eyes strains, and with all that, i still see a difference between the two pictures. (but i downloaded them, and switched from one to the other by zooming a bit).
I usually don’t see black compression issue on videos that i’m generating with VEAI (it’s a big problem, to see them, i must test each videos processed on my Tv (which has a bigger brightness without the LED “killing eyes” effect (old LCD Tv with a 16-235 color range), but still on this AOC LED monitor, i see the changes between the two pictures !
You’re making a another statement that doesn’t ring true. Studio drivers are best for content creators as Studio drivers have been optimized and tested with certain content creators applications such as Topaz products.
Game ready drivers are for gamers because there are many spot fixes for different games. Also you can’t record videos using game ready drivers.
You can use the version you want and the drivers are NOT the same.
AiDon: “the drivers are NOT the same”
I didn’t claim that. I just told my personal experience.
AiDon: “Also you can’t record videos using game ready drivers.”
The optional “Gforce Experience” tool by Nvidia offers video recording with both the game ready and the studio driver.
I love the fact that Photo AI (and Gigapixel) can seriously enhance these old family-type photos! I’ve been doing a lot of that – in my last blog I posted an example from an old tintype (mystery relative).
As for groups, if PAI or Gigapixel just won’t recognize everynoe, maybe try cropping the photo into sections and running each through separately? I haven’t tried that but it might reduce “confusion” in the app.
Other considerations: Scan the image at high res if possible, rather than photographing it, for the best detail.
Here’s a 2x treatment of your photo with Gigapixel, as an option (Mac version, Intel SSD iMac). It seems to have caught everyone in one go:
Please add an option to selectively disable faces for face recovery. I don’t want the pictures hanging on the wall behind my subjects to be face recovered!
Upgraded from 0.2, there are some issues seen, which were not present previously, I am testing it mostly on scanned photos:
no faces recognition at all after upgrade when using autopilot
when I manually enable face recognition, all filters are disabled on the image (even if they are enabled in GUI, and still no face recognitio at all) - just all filtering is off
I am using it on AMD GPU (RX5500M), maybe there is some licensing issue? I reseated it from my PC, because I wanted to test this SW on vacation
Hi ! Face recognition work here, but sometimes, it’s not. but it’s work. but i installed directly 0.4, not from a previous version, could it be related ? how many picture did you tried ?
i noticed as well the Gui issue you talk but differently. If i turn off the auto feature, all filters that were set up by the auto mode before work. but if while in Auto, i set up a difference Scale (like x2 or a 1080 height), then, the filters disapear when i click on Auto Off. Had to let the software do the auto procedure normally, then i hit Auto Off, and then, i can enter the scale, and all this is working ok.
(i tell this from memory of 2 days ago usage, maybe it’s a bit different, my memory is a mess )
I checked a couple of images, I have one processed and saved from 0.2 - and this looks like faces were recognised and redefinied compared to 0.4 output. Anyway I will try to reinstall - this is alpha, so something can not work Sharpen feature is very good, but it emphasizes scanned dust/speckles - but that’s side effect, maybe in future there will be AI engine which will remove dust without removing eyes:D
)
