Topaz Photo AI v2.1.0

GeForce Experience had not yet prompted me to update so I checked manually and see there was a new studio driver 436.01 as of 11/1/23. I did a clean install of that and ran two tests using a stopwatch.

Correction: Above driver version should read 546.01; not 436.01.

Test #1

Test #2 which I had hoped would be better was actually much worse.

Didn’t the ppl who prefer v.1 like the fact that they perceived it added less sharpening &, to them, looked more natural & traditionally photographic?

2 Likes

Exactly. In the images I compared I preferred the V1 in about 2/3 of cases.

Plus the fact that V1 leaves the brightness alone.
(The brightness issue was said to be solved, have to check…)

Thanks for the feed back. I am working on the CPU optimization.

3 Likes

All of this could be solved if they implemented my feature request: :grin:

2 Likes

Hi, i did a short seach on the nvidia site and the version i did find is: 546.01

Release 545 Studio Driver for Windows,
Version 546.01
Windows 11 / Windows 10

2 Likes

Good grief! No idea where I came up with 436.01. I am, in fact, on 546.01. At least I got the release date right. :flushed:

Thanks.

1 Like

Working on that. Thanks.

1 Like

Glad you are getting better results Ken. Hopefully I will too at some point as my desktop is also an RTX2060.

did install this one help?

I believe the call to remove v1 was specifically because the brightness issue was supposed to be resolved.

Here’s mine. Notice the protruding bumper on the driver’s side.

Bummer, the car picture that I posted was under a minute. Set to GPU.

1 Like

No, the results I listed above (3:53.56 on first try and 7:27.25 on second try) was with 546.01. I know I typed 436.01 but that was incorrect; it was 546.01.

1 Like

I got that on the first run as well, had to mask the wheel between the headlight and the bumper

1 Like

Thanks… One question:
Does TPAI 2.1.0 properly exploit the technology of the new AMD Ryzen 7000 ā€œX3Dā€ Desktop version, such as 3D V-Cache and the new AVX-512 instructions to speed up conversion, improving performance for the CPU-processed part? Can you officially confirm this for us?

Thanks!

3 primary Remove passes. Left corner shadow (5-6 secs.). Dark car (~ 20-30 secs). Orange truck (~ 60 secs). Sky replacement < 60 secs. Slight vignette < 10 secs. Reduce green saturation < 5 secs. Minor removes of ā€˜dings’ on classic car (< 15 secs.).

BTW, I didn’t do any denoising and/or sharpening to see what would happen with that b/c the interest seems to be in the remove…

1 Like

I am a little confused because, being a Beta Tester as well as a mod, I find that we are now using releases for continuing beta testing. It seems that Photo AI is a technology showplace rather than a robust image processing application.

I, for one, cannot use Photo AI for the inconsistent results when processing images. The application processes all images and especially makes random decisions about de-noising images with some textures such as stucco finishes being treated as noise.

We need to be able to set boundaries on the processing, especially for noise reduction and sharpening. For example, set a lower limit for inspecting and removing noise - say for images at above 800 ISO and for sharpening set limits for different types of uses such as for output media.

All in all this is a really rough ride with still using beta software for processing that uses expensive hardware. For example, one of the users commented that they process images using the remove options in seconds where most of us are taking minutes even with high end PCs.

8 Likes

This sounds like an interesting concept. What kinds of limits do you have in mind?

Is it different than the ā€˜tiers’ (or is it sensitivity - I’m not sure) set up in the Prefs? Is it something that goes beyond those parameters based on ISO?

And, for those limits, should it be applied only for Auto processing or in some way affect manual settings too?

It should be done for all images/processing to hand some control back to the users. The Automatic processing is inconsistent across image types giving different results. For example, I can export a TIFF and PNG and have different processing applied automatically. RAW images are even worse as, apart from ISO, cameras are sensitive to light and light direction such as backlighting etc.

It is easy to test by using studio lighting in different ways, shoot something on a textured background etc., etc. AI is a dumbed down version of thought and observation that doesn’t take artistic impressions into consideration.

2 Likes