Topaz Photo AI v1.3.12

Hello!

This week we have a handful of updates related to different cameras as well as some fixes for some annoying issues.

For general fixes, there should no longer be a large movement in the preview when zooming in/out for the first time. We’ve also fixed the Open Recents menu staying open after selecting an image on Windows, as well as adding an additional option to clear all recent files.

For cameras, we have added profiles for:

  • Fujifilm X-S20
  • Leica Q3
  • Sony ZV-1M2

We have also updated profiles for:

  • Canon EOS R50
  • Canon EOS R8
  • Google Pixel 2 XL
  • Google Pixel 4a Rear Wide Camera
  • Google Pixel 7a
  • Nikon Z 8 (Does not fix unsupported image issues)
  • Samsung Galaxy S23 Ultra Rear Super Telephoto Camera
  • Samsung Galaxy S23 Ultra Rear Telephoto Camera
  • Samsung Galaxy S23 Ultra Rear Wide Camera
  • Samsung Galaxy S23+ Rear Telephoto Camera
  • Samsung Galaxy S23+ Rear Wide Camera

On top of these profiles, DNG files with missing profiles should now no longer output invalid DNG files. We have additionally updated lensfun definitions to be up-to-date as of June 23rd.

As always the full change log can be found below.


Please give us any feedback or report issues with this release. We’ll be updating TPAI regularly to address those pieces of feedback and issue reports. If there’s a specific image you’d like us to see, you can send it us at this dropbox link.


v1.3.12
Released June 29th, 2023
Mac: Download
Windows: Download

Changes since 1.3.11:

  • Updated lens correction files
  • Updated camera profile support for various cameras
  • Added fix for unsupported DNG files not creating valid DNGs
  • Added Clear Recent Files option under File > Open Recent
  • Renamed Detail to Refine Detail in Remove Noise and RAW Remove Noise
  • Fixed crop on certain RAW images
  • Fixed images with invalid orientation causing issues
  • Fixed zoom not being centered on first zoom in/out
  • Fixed Photoshop plugin not correctly storing settings
  • Fixed open recents menu staying open after selecting an image on Windows

Known issues:

  • HDR or Monochrome DNGs which previously did not open may have issues when saving as a DNG
    • To get around this you can save as JPEG, PNG, or TIFF which shouldn’t have the same color issues
12 Likes

When will batch processing be available?

Keeps getting better. :slight_smile:
1.3.12 seems so much faster than the previous versions.

Still hoping to see support for the M sized compressed RAW Sony A1 files. I guess I can process the RAWs elsewhere and then run them through PAI until then.

Hello,
thanks for the good work. Will there be updates to batch processing and CLI features in one of the next releases? (I am hoping for that every release, but the focus seems to be on smaller bugfixes, cameras and models).

Batch and CLI in their current form are not usable in a professional way when working with hundreds or thousands of images.

  • One cannot set what models to use for Batch Processing Autopilot (or alternatively disable specific models). Most of the time when batch processing one knows exactly what kind of images are going to be processed (computer generated, photos, and so on - but there is no way to specify that; it just autopilots over it…)
    See Additional preferences for autopilot and Is there a way to disable models?

  • One also cannot really batch process using the “Apply to all” feature, as it doesn’t only copy the selected models, but also the image specific values it detected for the first image to all other images (instead of re-detecting those values like sharpen strength / sharpen clarity / enhance suppression noise, … for every image)
    See How can I apply current settings to all loaded images without their values?

  • Since “Apply to All” is not really usable, one has to click on every image manually and change the models. But the film strip UI is not build for that in any way (if you have hundreds or thousands of images you can forget it). Gigapixel had a much better UI for batch operations and multi-value changes:
    See Port the spreadsheet style interface that gigapixel had as a UI option

  • The CLI is unusable as an alternative, as it does not allow specifying models or options.
    Instead it will autopilot every image and then you’re no better off than having used the UI.
    See Photo AI CLI - allow arguments to influence settings

Will any of these issues be addressed in one of the (near) future releases?

Thanks in advance!

8 Likes

Hi,

Thank you, it seems that .12 release fixed the instability of previous release (that was not after all NVidia driver issue).

However, it seems that .12 has somewhat increased tendency for artifacts (see attached image). This is from the Standard sharpening model and occurs with practically all stength and clarity settings (crop from the bokeh area of image, ORF-file).

1 Like

This release and the previous few seem to me to be retrograde steps in terms of colour. You almost had the RAW conversion producing a Gretag chart that was usable, but recently it has returned to poor.

What became of the additions promised for release many weeks if not months ago , the rotation for crop to correct horizons ? Seems to me that at the time that you removed the option to upload our images to your development of AI , the addition of new abilities ceased.

I can confirm that this new version has indeed fixed the violating zoom bug when zooming in on a new image for the first time. This version also seems to be a little faster than the preview. Another very good point. I haven’t tested all of this new version yet, but on what I have already tested, it’s not bad.

1 Like

It would be pretty cool to be able to have some preset options that one could use for a batch process.

I can see shooting an event and having several hundred shots to process at once with a similar look or one preset for the reception / another for outdoor shots / etc. all ready to go with a click.

1 Like

Bug Report:
Loaded image1, adjusted settings
Loaded image 2

Trying to select Image1, so I could apply it’s settings to image 2

BUG: Image 1 cannot be selected!

What is actually happening when I am clicking on image1:
image

However, in preferences, subject detection is disabled:

Subject detection does not seem to finish at all.
Using M1 Max

Feature request: images captured using Nikon 200-500mm with teleconverter have a bit of a halo. This is rather a common use case, that teleconverter leads to softening/haloing compared to the lens itself. This is most prominent at the edges with bright highlights. Would be great to have the ability to correct for such effect.

Before, 400%:

After, somewhat the best I could tune, but halos are still there - around the beak and head the bird’s chest:

there is already some sort of batch processing.
But I would prefer it to allow to copy the current parameters to only a selected pictures not to ALL pictures.
The fully automated batch is still not very reliable.
They should try to reduce the noise variably on different part of the image. The level of noise in the pictures are different (higher in the darker areas)

Low resolution model create lots of artifacts standard model works fine.

I keep getting “internet connectivity error” while trying to install, what could be the cause of this?
image

It complains about low disk space on C:, but I had previously (IIRC) set both the Data and Temp directories to a separate HDD and SSD respectively. However the option to set these doesn’t seem to be there any more - why? Can we have this as an explicit thing, in setup for the data and the program for the temp files please? (Or am I missing something?) I’m sure (see below) the models are still where I want them, not sure about the temp files?
Notes - I do have:
PAI_MODEL_DIR=E:\TopazData\Topaz Photo AI\models
Which only exists while the program is running, and seems to drag in crashpad and logs at the same level.

Despite all the updates to TPAI, I still think that the original softwares often do a better job, even though they have not been updated in months.
I took another look at a shot I took a few years back, where the foreground rocks were hopelessly out of focus.
In both cases, I used the out of focus model cranked to the maximum and also had to spend a lot of time refining the mask. This was slower in TPAI, since there is no edge aware tool and the AI brush couldn’t handle the transition between the rocks and the waves on the lake at all well. In fact, in both cases that was the most difficult refinement to achieve.
Even in comparison view, one can see far more detail in the SAI version on the left. The TPAI version also has a pink artefact.


Zoomed in, one can see that there is far more blur and blotchiness in the TPAI version:

The original DNG was processed in the standalone version of both softwares on a Mac mini with 16GB Ram and running Ventura 13.5.

3 Likes

Hi
Today I refreshed my Topaz Phot AI to v1.3.12, but the result is strange.
I seems the system doesnt work or something wrong with the preview.
(But I made processing and it doesnt finish it after several try)

Thanks for your kind feedbacks…

Cropping seems to produce sized down versions of sharpen areas as artifacts in the resulting image:

Here’s the RAW file (Sony Alpha 1 Lossless Compressed RAW):

I can confirm this does not occur unless I crop in TPAI.

When I send a photo to Photo AI it arrives in AI without the Lightroom edits. For that reason I’m finding it unusable at the moment unfortunately.

You can sync the DNG returned with the original or use the Edit in option and send a TIFF and include the edits you have made.

Other applications cannot read and apply LR edits as the have different RAW processors.

Hi Don,

I use edit in Topas photo ai which normally sends the LR edits to TPAI. But for some reason it has stopped taking the LR edits with it.