I like the improvements. Using it much more as my go to de noise app. But why on earth can’t it automatically recognise a landscape. Instead it picks a seemingly random rock or tree.
1 Like
I just upgraded from 1.1.9 to 1.2.2.
I ran 2 images through the software and got marginal improvements. These seemed to be images that would have been slam dunks in the past. Ran one through Sharpen AI and got much better results.
Could the upgrade have failed? I didn’t receive any messages.
Fun fact: Here’s 1.2.2 running on a 2013 MacBook Air (with unsupported Monterey 12.6.3)!
It’s painful slow but it works 
I have over 10 Topaz software products starting with Studio 1. I now have Studio 2. It seems that every time I get a new Topaz product, Topaz stops upgrading previous products, kind of like bait and switch. How about upgrading Studio 2, which is falling behind Photoshop Elements (which I have)?
Early on in the Photo AI development process I suggested they simply continue developing Studio 2 (the foundation for moving forward was already there), but they explained Topaz is getting away from “artsy” effects in favor of corrective treatments.
The good news is, Studio 2 still runs (at least on Macs) under the current OS, so it’s still viable and usable.
1 Like
Studio 2 was a free upgrade to Studio 1, both are now no longer sold or updated and regarded as legacy product. You can still download the last version from the legacy products page.
1 Like
Win 11 desktop PC. Ps 2023 Plugin. Processor = AMD RX6800 XT.
I don’t remember PAI plugin being this slow to load (min. 20+ secs) an image from a Ps layer. Or, take as long to generate a preview (12-15+ secs) since the alpha days…
Something seems very off.
The plugin product launched virtually immediately from Ps in all the recent several months releases. And in the past couple months, the preview generating speed had been getting exponentially faster. Not this release! Certainly not this run of it.
This felt like a major step backwards wrt speed of ops.
The sharpening it did on the image I used I think did a realistic, natural job. Snip attached. It didn’t over-sharpen and/or create halos. I enhanced what was produced with some Structure slider adjustments in Nik Viveza (also as a Ps plugin). But I typically do more processing on all my Topaz plugin output layers. That’s why I don’t want pixel shifts, etc. or anything that will be aggravated by further post-processing in Ps or other 3rd party filters…
BTW, this jpg snip looks softer here than it does on my monitor. I chose this to test b/c of all the fine details on the bloom & branch. And, to use something other than a bird (which Topaz seems to like to use to prove product efficacy).
I just ran an image as a layer from On1 PhotoRaw 2023 to v.1.2.2 and back again on my Mac mini M1 16GB Ventura. TPAI was on Autopilot default, ie with face detection on, although the face was a monkey’s. Auto selection was not accurate but I didn’t refine it. It took about 40 sec to send and and generate a preview and another 12 to save and send back to PR. I cannot recall how this compares in terms of speed with earlier versions of TPAI.
Sharpen AI was quite a bit faster on the same image.
Win 11 Pro desktop PC. PAI 1.2.2 Plugin (Ps 2023, rel. 24.1.1). PAI Processor = AMD RX6800 XT.
.ARW orig image opened into Ps 2023 via ACR (no adjustments in ACR). Image launched from layer stack into PAI plugin.
This image behaved more like how the past couple months’ plugins were wrt launch and processing time. About 4 secs to open from Ps into the PAI plugin. And processing time was faster than last night’s tree image too - it went relatively fast until I turned on Face Detect; it needed a few moments to regroup itself and then kicked in.
I don’t typically ‘shoot’ people. So always have Face Detection set to Off by default. This group shot of the Stanford (Univ.) band (sorry, no tree in this image…) has a lot of small faces. I turned Face Detect on manually. It said it was able to detect 18/19 faces.
I understand why it wouldn’t pick up girls with hair or pom poms across their faces… but PAI Face Detect didn’t get the girl whose face is fairly clear at far left (with the sax) nor the guy just above and to her left (camera right - with the trumpet). They are in a bit more shadow than the others, but should still be detected. There are evident facial features on them and they’re in the same focal plane as the others.
I forget if there’s a way to add missed faces, vs just turning on/off detected faces. Is there?
Was able to save back to the Ps layer okay. Slight pixel shift (down & to the right) on the returned layer. Able to do add’l post-processing in Ps to add exposure to the subjects in foreground and desat & depth blur objects in background. Interestingly, Ps “Select Subject” didn’t like (or pick up) the girl with the sax at far left either (perhaps all these AI robots are primarily acoustic fans and not brass???). I didn’t add her to the set of subjects (purposefully) so you can see how she was treated differently than the others wrt expos., vibrance, etc. I was going for a somewhat dimensional look vs 2D flat look to the processed image.
Before/After PAI snips (in the PAI UI) & Ps layer stack w/PAI output just above Bkgrnd Layer (labelled in the stack) snip.
It has been requested:
Add ability to select faces that are not recognized by Photo AI
You can add a comment and like to it.
5 Likes
Thx! Just upvoted.
In my hierarchy of needs, the Manual Masking features I’ve previously requested/articulated in Dec & Jan Roadmap threads would have much higher priority. But it would be useful to be able to ID add’l faces in some instances when desired.
1 Like
My patience is gone. Can I have my money back?
This is a user to user forum, no one here can give you your money back so you will need to raise a support request at the main website
Stopped using PAI for a while as the denoise was just too strong.
Tried v1.2.2 and am pleasantly surprised.
Have changed my workflow so that images are edited in LRC (except for denoise and sharpen) before exporting via plugin to PAI.
There is still some work to do to improve both sharpen and denoise in PAI but the progress this far is heartening.
Thanks 
There is a huge uptick of Mac downloads with 1.2.2 and it is only 3 days since it was released. Currently 2.3k downloads, but for previous versions it was usually 200-400 even after a week or more. I imagine the Topaz marketing started promoting it more on Mac websites or something. The better Apple Photos compatibility seems like it got some good word spread around.
Do you think it could it be a rhetorically question - just to show my opinion about a ongoing project, where nothing of importance moves.
I can only urge this development team to find out how DxO have made their AI model - or buy a license, before a lot more ex-loyal customers lose their patience.
Everyone is entitled to their own opinion. I don’t agree about DxO as I think the DeNoise models are far superior.
I also don’t think all noise needs to be removed and unless noise is really bad, usually color noise, I don’t bother as noise removal destroys the structure and creates a false look.
2 Likes
I believe in this case it’s because the devs asked us not to upgrade from within the app, but to download anew.
2 Likes
But we can’t take Denoise in PAI to zero… And, it sharpens too.
2 Likes
in photo a1 1.2.2 i am getting pink artifacts when photo AI does default processing and result
