Looks great, but the macOS version is broken and will not replicate those results. PhotoAI face recovery does not work at all on my system. It helps with other image improvements, but I got it specifically for the touted face recovery.
It’s gonna be fixed in this Thursday’s update
Unfortunately we are at version 1.3.4 and are yet to have a stable release with all options working at the same time, even the basics, on both Windows and Mac OS.
It is a shame that there is no confidence in Photo AI but, at the same time, there hasn’t been any progress with DeNoise AI, Sharpen AI and GigaPixel AI since Photo AI was released.
7 Likes
But what will be the new bug regression that arrives in the update? 
Today I saw the video to version 1.3.4. I am looking forward to use the new functions. I’ve tried to activate the new denoise functions. But there was following message: “COULD NOT LOAD MODEL. OPEN SUPPORT ARTICLE”
So my guess is that the new beta functions caused the errors.
Did Windows deactivate the new beta functions during the repairs (Restorepoin)t?
It is true that for a while now, some bugs have been corrected, but others that didn’t exist before are occurring. It is quite disturbing.
1 Like
I open a TIFF file (processed from Nikon NX Studio V.1.3.1) as soon as I open the TIFF file in Photo AI v 1.3.4 after subject detection it crashes.
2023-05-18-11-05-33.tzlog (74.7 KB)
I tried few times already. Already rebooted the machine (windows 11) and same problem.
Attached the log file for the same
I do not even touch the mouse or the key board after Photo AI and selecting the TIFF file.
had 1.3.0 for couple of days, worked like a charm, got upgrade offer to 1.3.4, upgraded,
all Recover Faces results are awfully blurry, when processing good, sharp photo the results visible in preview are completely unusable.
Recover Faces function is totally messed up.
Reinstalling 1.3.0
I seem to be having the opposite problem to what many are describing:
In the preview, face recovery works fine, but when I open the resulting .jpg file, it’s not as finely detailed as the preview was.
Preview:
Actual opened .jpg afterwards:
Have tried several times now.
Recover faces is not intended to work on larger or full frame faces. I believe the explanatory text in the Settings panel mentions that…
What happens if you just use Sharpen settings?
P.S. I know it’s not clear or intuitive from the design of PAI that there are Tool Tips available. But if you turn on the Preference to allow mouse hover over pop up instructions under the Edit menu, there are brief comments about usage of the models/tools. They used to include little ‘?’ symbols to give folks a clue there are tips available…
Hello,
I have worked with every version of Photo AI so far without any problems.
Version 1.3.4 is the first version that crashes.
It is sporadic, not necessarily traceable:
- often at the end when several files have been processed.
- sometimes directly at the start when the first file in a row is to be processed.
The program exits without an error message.
Settings:
AI Processor is set to Auto
Autopilot, Subject detected, Face recognition is off,
Remove noise or Sharpen normal, Preview was calculated
Save image in overwrite
Program crash without an message at the start or the end of the save image process
Computer:
Windows 11 Pro build 22621.1702
i7-8750H CPU @ 2.20GHz 2.21 GHz
32,0 GB RAM
NVIDIA GTX 1060
Edit:
I assume that the problem is to be found in that direction:
1 Like
Before posting a comment like this one, be sure to read the previous posts. The problem of “Recover Faces” is known in this version. It should be fixed in the next update which should be released today at the latest.
This seems to be getting worse all the time. As of (IIRC) 1.1.6 it upscaled better than GPAI, but this hasn’t been true for many versions, even despite the dreadful new standard model in gpai 6.3.3. Can we just have a stable version designed to work reasonably well and move the “we break stuff every week” versions into a beta program (and if a good one turns up make it the new stable version)? Or is it all so flaky there is nothing stable available?
How about a system of voting for the best versions of all your current products. The ones with the most votes get to be the stable versions and everything else is a beta… For example I’m going with:
Video AI 3.0.9
PhotoAI 1.1.6
Sharpen AI 4.1.0
Denoise AI 3.7.2
Gigapixel AI 6.2.0 (but 5.4.5 was great too)
1 Like
It’s an interesting idea.
But, of course, everyone can just treat all new releases as public betas - since that seems to be the tech trend (generally) in the past 5 yrs to push out 1/2-baked products & fix them as users complain. There’s even a term for it, which I now forget.
Someone somewhere thought it was a smart, competitive market strategy & companies jumped on that bandwagon…
So, to your point, assume new releases will be buggy. Watch & wait to read the comments of those willing to be public, bleeding edge, beta testers & then decide whether to leave the comfort of your stable release based on relevant comments. I do that for Adobe all the time given how they’ve been going, so it’s relevant here too…
BTW, what is it specifically about your upvoted rel. that makes them your faves (not just “stability”, which is not a measurement qualifier)?
Thanks, but I don’t get why the resulting saved .jpg has far less detail than in the preview window? And it’s not due to .jpg artifacts.
Open the images I posted and zoom in on the beard. The preview shows much more detail / fine hairs than the produced .jpg.
If the software is capable of the preview quality, wouldn’t it be fair to say the saved image should be of that same quality?
Yes, I agree with you on that - not just applicable to your face example either.
Plus, you didn’t have other tools turned on (at least not denoise & sharpen) so there should be no excuse that the sharpening you were seeing was from them. I’ve experienced the same thing in the past for other types of images - since I rarely shoot portraits and less often than pre-pandemic shoot street photography.
Agreed it’s basically pointless to have a preview showing you what your image will look like and have you spend time pushing sliders around to get the preview to look just right (to one’s taste) then hit “Go” and it becomes something else. The whole point of a preview should be what you see is what you get. (But, that said, you do have to use the right tools for the uses they were designed for too… 
).
JPG is a lossy format so that is not a good test that saving is losing quality on faces, use a non lossy format to save to prove that. Even if you have adjusted jpegs for little compression highest quality, jpegs can have less color resolution than grayscale resolution.
