Just FYI if you search for support on this forum you will get the following description.
If you need to report a problem use the Support page on the main website, and, just for your information this is a user to user forum that does accommodate some error reporting on a general basis, not one on one:
And on the website pages you have this callout panel, where there is a option to click on “Email” or “Chat” online during working hours:
How do you feel about the result of this image? What you think about your image is what matters most …
But, to my taste, the face looks as if it was cut & pasted on and is (to me) way too sharp and distinct relative to the quality of the rest of the image so doesn’t look quite right.
If it were my image (and, yes, I know it’s not…) I’d take down the contrast on the face (b/c the hair and neck softness, as well as the clothing, don’t sync well with the face). Then mask the face (so no more processing could be performed on it) and see if I could get the hair, neck, clothing to be a closer quality, sharpness, contrast to the face. And, I’d leave the stone wall in the background soft. Is that doable using a mix of Ps & PAI?
Based on my approach style - this is what I came up with using PAI 2.3.0 - including Subject detection, denoising, sharpening and face recovery. Then a hint of Ps neural filter Photo Restoration (b/c I’d done the face recovery in PAI I pulled the face recovery slider in the neural filter to the far left). My goal was to try to balance the sharpening and contrast as close to uniformly as possible on the subject (I masked out and didn’t touch the background). The image isn’t sharp sharp, but I went for consistency of sharpness & contrast… Anyway, it was a fun Ps challenge to see what would result! I think the face might still be a tad too distinct in mine - perhaps it’s some toning that’s needed to match better with surrounding content. I forgot to mask the b.g. in the neural filter…my prior step looked better on the wall. That could be masked back in.
In my humble opinion, you must be doing something wrong, because to be honest I’ve never encountered this kind of problem since the first versions of the software. It downloads all models once during installation. And that’s all there is to it. It doesn’t download them again.
I find the additional processing in Photoshop using neural filters unnatural, unlike the first pass in photo AI. It’s far too sharp compared to the rest of the image.
Maybe send it to Stable Diffusion to reconstruct overall image detail.
OK, seems like I’m too stupid for Photo AI. Then please explain to me what I’m doing wrong when I call Photo AI via the Filter Menu or Automate. What secret have I not yet discovered? And what about the other users reporting this problem?
The fact is that the problem did not occur initially. Everything was OK until version 2.1.x and not anymore since 2.2.x. Every time I do an update, all model files are downloaded, so you can assume that everything is OK. Nevertheless, the files are downloaded every time.
I basically have the same configuration but I have 32 g and I have the same problem it can’t recognize faces just spins
Processor Intel(R) Core™ i5-6600K CPU @ 3.50GHz, 3501 Mhz, 4 Core(s), 4 Logical Processor(s)
Total Physical Memory 32.0 GB
Installed Physical Memory (RAM) 32.0 GB
What GPU do you have with dedicated vRAM and the driver version?
Adapter Description AMD Radeon RX 6600
Driver Version 31.0.23013.3005
see attached file
(Attachment Image1.tif is missing)
I have the same problem with most images it just stops at “detecting faces” with autopilot continuously running. I also have a similar problem with Gipapixel AI 7.0.1. However, Denoise AI runs fine!
Windows 10 & Nvidia GeForce RTX 2060 Super
Sorry it didn’t like my screen scrape
Adapter Description AMD Radeon RX 6600
Driver Version 31.0.23013.3005
total available graphics memory 24554 MB
Dedicated video memory 8192 MB
Shared system Memory 16362 MB
I’ve never tried stable diffusion. I’ll have to look into it. I assume it’s some kind of AI program? Thx for the pointer.
This is a Photoshop Beta using the Restoration Neural filter as is. No adjustments, using default settings. Consistent focus throughout. Just as good as TVAI in my opinion. No patterns on the dress and background. Plus heavily JPG compressed on save.
Isn’t it in the Preferences anymore?
There used to be a way to turn on (or, off) Lens Correction there.
Thanks for the reply!
I don’t use Ps beta. Always interesting to hear what ppl do with it.
I seem to be having a few issues with the lens correction thingy.
If I process a RAW file via Lightrooms file/plug-in Extra/PAI… w/Lens correction turned OFF in PAI, it returns a DNG with the original lens correction profile from LR… which is what I would expect…
If I do the same thing except use the Edit In drop down directly from within the image, and send PAI a TIFF file, the returned TIFF file’s lens correction profile from PAI is stripped.
Yay! Glad you found it.
Have a look at this report on the forum where AMD GPUs were having was having issues with Photo AI and the OP downloaded a fix from AMD forums …
Lens corrections are only applied to RAW images so apply them before you pass it. Note this applies to all RAW processors.
That’s good to know!
I’d turned it off in any case in my Prefs. But since I use PAI via Ps it wouldn’t have been relevant in any case (& therefore, IMO, better to have it off to prevent any processing issues).
