Topaz Photo AI v0.9

Still don’t understand why sharpening is so bad in PAI while you proved to do it right in SAI? Why don’t you use the same engines as the standalone programs and melt them into 1 single program?
Here you can compare what autopilot did, what sharpening alone did over the whole picture without noise reduction and what SAI made of it. I prefer SAI!



1 Like

In my case is still crashes both as a standalone and as a plug-in.
Unlucky me; cannot even test it.

Still only 8 bit tiff for RAW files. Tried a png but it also seems to be 8 bit and sRGB (according to Affinity Photo). Affinity photo says no camera data for the png. But sending file from Affinity Photo went well. Sending a tiff from capture one seems to have come back as 8 bit but the original colour space. DNG from RAW seemed OK.
I have 2 monitors and would like to specify which one it opens on. Is that possible? On one it fits the screen OK but the other its too big. And I like to keep one screen for my main editing program and the other for plug ins. De noise, Sharpen and Gigapixel always open on the correct screen. Anyway for me to use it more often I need 16 bit. But I still like it.
For RAW adjustments I’d be happy with a histogram and levels so I could set black and white points.
It uses a bit of memory!

Oh oh looks like the cat melted :face_with_hand_over_mouth:

1 Like

I have the latest PS version 23.5 & yes, I duplicate the layer every time.

Hi, If I process a raw file via the new RAW Plug-in from Lr, Photo AI exports the resulting DNG file back to Lr normally. If I send the same raw file to Photo AI again - to try a different approach (e.g. one with/ one without sharpening, so I can then use a mask in Photoshop to refine the results) - the DNG export from Photo AI produces an internal error in Lr. It appears an ‘overwrite or use unique file names’ pop-up needs to intervene.
(I guess this is also another way of saying it would be nice to be able to refine the mask in the sharpen module - though I appreciate you’re trying to keep this a cut-down/ mainly auto app).
Otherwise, it’s producing some very nice outcomes (for my preferences I generally have to reduce the auto settings - particularly so with the sharpening).
(Raw Files from Canon R5. MacBook Pro M1 Max, OS Monterey v12.5.1. Lr v11.5. Photo AI v0.9.0).
Phil

I am still unable to find a plug-in for Photo AI in Photoshop Elements for Mac. I only see plug-ins for Sharpen and Denoise. The menu bar does not give me an option to install the plug-ins and I can find no way to do so manually. Wasn’t this something “corrected” in version 0.7.2??

See screenshot.

Glad to see the resize buttons back, thanks for that.

However Autopilot seems pretty broken now. Remove Noise is being triggered for almost every photo and just destroys the image, turning it off and using just Resize does better noise reduction. It also seems to hardly ever detect faces any more.

Trying to find images I have done before to see if its just having issues with the ones I just did.

1 Like

Similar issues here when upscaling old scans etc. I am finding I am getting better results ignoring the Remove Noise options entirely.

2 Likes

Its not that Remove Noise was terrific before, DeNoise itself was always a tool I used sparingly as Gigapixel did a better job for most photos. Its the fact Autopilot is really aggressively turning it on when its not needed and produces an unusable result compared to turning it off that often will produce a perfect result.

I hope that’s the case. I am reworking some old scans and the current version of GPAI keeps crashing so is near useless at the moment for me.

This is great. I installed v0.8 yesterday for the first time and enhanced a portrait this morning, but the face (which was a large part of the image)wasn’t detected. I saw v0.9 was available, installed, and not only was the face detected, but it did an incredible job of enhancing.

I love the simplicity of this new version, but it does feel like there’s too little control. In Gigapixel I could choose to turn Face Recovery on or off, but if a face isn’t detected here, then I can’t force it to be turned on.

Autopilot is also incredibly impressive, but again, it seems to take over no matter what. I’m hoping the final release will have the option for more manual overrides. Or perhaps whatever changed in v0.9 will make the edge cases where Autopilot doesn’t work as well, less likely. I see that I can override any of the automatically chosen settings in the filter options, but there are fewer options than in the previous standalone apps.

Lastly, it would be great if we had some Subject Mask Refinement options. My subject was detected well, but the masked hair was very blocky, which means sharpening the subject only looks terrible around the hair. The ability to blur, contract, expand, and even paint a mask would be welcome.

Fantastic tool, everyone! I love the combination of all the tools from the IQ bundle.

the “apply settings to all” function is still missing since 0.6

Note how the result with remove noise applied (on the left) creates a halo around the moon and the horizon. The version without is much better in my view. I note too that there are more detailed suffixes than there were. But shouldn’t “enhance” be “resize”?

Download of 9 was slow, must be a big update this time !
As soon as it loaded noted the removal of the RAW adjustments, and the move back to a trimmed down selection of variables. I guess this begs the question DXO 5 or Pure RAW 2 ?
If it is heading towards the latter than I note that the selection of subject ( Porsche on road ahead of white car , same old shot) now chose the Porsche as the subject correctly in my mind, however selected the car and left mirror but not the right ! So if it is to be an AUTO program , how do I recover a sharp left and unsharpened right mirror. Again in the AUTO selection process the car is sharpened ( great ) and noise reduced, but the background is neither . So I now have a part sharpened car and a noise free car against a noisy background. This is alleviated by moving to control and selecting noise reduction … without changing the numbers chosen on Auto the car and background are now noise free. So I need to take it off auto to get a workable choice.

The RAW conversion is acceptable now , be nice to have a slider for minor changes but these would have to interact with the display image in real time. ! For me , I feel my most preferred choice would be the ability to adjust the paint selection mask , and have some RAW adjustments , as well as the ability to decide on the amount of noise in the image. I fed it a copy of an old Kodachrome and it said medium noise ( this image has loads ) , prior version of this allowed weight of noise adjustment for heavy and to my mind worked better.

So if it is a Pure RAW v DXO 5 decision circulating, I might pick something between , less than a big program and a tad more user selectable than a one stop shop. PS Windows protocol on window area used is still not observed and it is annoying for screen allocation when dragging and dropping from bridge or other image tool. If it becomes a simple drag and drop one stop … may I please continue to have updates for my Sharpen AI and Noise AI and plugins for Photoshop. ?

I wonder if NEF is the issue…

How large are your original files? Are you using medium format camera then doubling the size by duplicating the bkgrnd layer? If so, wonder if it’s overwhelming PAI.

I’m using 23.5 on Win 11 and it works.

But ever since I was told the raw files aren’t sent to Topaz products via the plugins I don’t try to process anything but tiff and jpg and png (from Canon, Sony, Olympus originals)…

Marked performance difference from earlier versions…congrats!

Pros: Autopilot made better decisions about noise reduction and sharpening, both of which were backwards in previous versions. Really like the RAW workflow improvement with Lightroom on the Mac.

Cons: Issues downloading AI models as noted by others, but doesn’t seem to affect usage. Significant exposure changes when coming back to LR from PAI as DNG. Some noise reduction occurring even when off. Initial photo scan seems quite a bit slower than previous versions.

Isn’t that the subject detection on the cat with the red overlay?

Wouldn’t it be a thin yellow box just around the head in the center workspace for face detection?

If not, perhaps you should call the cat Dr Jekyll…

1 Like

Yesterday I took several photo’s of students sitting in rows.
As the light was not optimal and I could not use a flash a lot of them were noisy.
I needed to denoise 40 photo’s and have them ready before noon.
So I decided to load these 40 in Topaz Photo AI for batch processing.
Apart from removing the noise Photo AI also decided that Face Recovery was necessary.
Most photo’s did indeed look better but the face improvement did not take focus depth in account.
So some faces in rows further away were also “recovered” while the persons next to them not.
I do not know if there is a solution for this but I just wanted to let you know.
The photo’s are for the school (I am a grandfather) and for privacy reasons I cannot include any, sorry.

Best Regards,
Luc

1 Like

Major performance improvements in 9… Great job @anthony.lawn Exporting RAW images via the Plug-in Extras is near instantaneous… The Face recovery thing is creating a serious issue with an exposure box around a face… Any DNGs returned have a 2+ stop exposure box around the faces…