Topaz Photo AI v0.6

Hello everyone!

This one includes a few big additions which are DNG export and the ability to use the subject mask to mask Sharpen output as well as some smaller things like autopilot tweaks and of course bug fixes.

For DNG output if you notice any major color shifts please let us know. There is still a lot of work to do in regards to perfect color accuracy so there will be some slight shifts, but anything where it comes out say green/purple we will prioritize looking into.

If you own DeNoise, Sharpen, and Gigapixel with up-to-date licenses then you have access to Photo AI. You will need to claim a Photo AI license by going to your account before you’ll be able to login.

Released August 11th, 2022
Mac: Download
Windows: Download

This is an early access release, so there will be some issues you may run across. Issue reports and general feedback are much appreciated. If there’s a specific image you’d like us to see, you can send it to us at this dropbox link. Below the changelog are a list of issues we’re already aware of.

Changes since 0.5.0:

  • Added ability to export as DNG
    – Note: This only works with DNG or other RAW files as input. If you attempt to save a non-RAW file as DNG it will give you an error on export.
  • Updated Autopilot logic to account for luma noise and any applied resizing
  • Added ability to use the auto-detected subject mask to mask Sharpen filter
    – You can toggle this on and off inside of the Sharpen filter settings
  • Added ability to close a single image from the UI (bottom left, left of progress bar)
  • Added a warning prompt when exiting the app with unsaved changes
    – This does not handle every edge case and does not handle closing images individually, but will be - updated soon
  • Updated Exiftool to 12.44
    – Should add support for some newer camera metadata
  • Face recovery will now be locked/disabled when no faces are detected
  • Changed Close Image and Close All Images options so that they don’t show up if they would close the app (external editor mode)
  • Added a support link to the Help menu
  • Added the Photoshop plugin to the install directory on Windows to allow people to copy the plugin to other apps even if they don’t have Photoshop
  • Changed login dialog buttons and error text slightly
  • Changed Autopilot wording for some cases
  • Added a blocking overlay on the right panel while authorization is being checked
  • Changed face detection rectangles style slightly to be clearer to see
  • Added more status messages when loading an image so that long images no longer just seem to hang in one place

Fixes:

  • Fixed menu items on Windows being empty in plugin mode
  • Fixed Windows app not responding correctly to minimize/show events when clicking the icon on the taskbar
  • Fixed a potential download speed issue in the installers
  • Fixed plugins not overwriting correctly in some cases
  • Fixed a model erroneously not being downloaded on Windows, causing a No Filters Active issue

Known issues:

  • DNG output cropping may not be correct
9 Likes

You stated that version .6 has the “ability to use the subject mask to mask Sharpen output.” I am an aviation photographer and like the propellers to be blurry to show motion. In Sharpen, I am able to subtract from the subject mask but I do not see that option in Photo AI.

2 Likes

That is correct. For now there is only the toggle to sharpen only the detected subject, or the whole image.

I think the crucial words are the “auto-detected” part of the ability:

Added ability to use the auto-detected subject mask to mask Sharpen filter
You can toggle this on and off inside of the Sharpen filter settings

It would appear the ability to modify the auto-detected mask is not included. Just the ability to toggle the auto-detected mask on or off.

Will you be adding the ability to add/subtract from the auto-detected mask or be able to create your own mask?

1 Like

Things may change in the future, and decisions are being made and re-made every week, but for now I think we’ll be seeing how well the subject mask does, and working to improve its accuracy.

1 Like

When multiple images are loaded, selecting manual settings on the first image (I tend to use scale 2x and enhance-natural) and then selecting “use these settings for all” does not seem to work. All images past the first one run the entire autopilot logic on them (taking tons of extra unneeded time and processing) and then apply the autopilot determined settings ignoring the manual settings from image #1. The only manual setting that propagates to the images after the first one is the scale factor.

1 Like

Please lock the v0.5 release thread

I put a timer on it a few hours ago just so people who had anything they wanted to post there can wrap up. Should auto-close in about 20 minutes.

1 Like

Thanks for the update … just tried it in C1 22 (latest version) and was wondering if it is possible to bring it back into C1 as a .dng. Right now it comes back as a .tiff file.
If possible how do I do this?
Thanks again
Ron L

On the only picture I checked the mask can be very good it detected two people with Green countryside background. But the mask is changing with the level of noise on the same picture less noise better mask. Nevertheless final result with face recovery and Sharpen with mask very encouraging
Still some eratic comportement with lost of control of the interface , need to close pictures to recover the normal interface

Just a quick test on this new version & i’m seeing pixel shift to the right and down when sharpening is applied.

Comparison of Images
Nikon D850 ISO 8000
Left to right - Original - PAI - PAI+Sharpening - Image processed with other NR program prior
The original pass w PAI controlled the noise, however it also significantly reduced the detail in the fur. When processed using PAI+Sharpening (Default lens Blur), artifacting in the fur occured. Included is a prior pass through another NR program that I had used on this image prior for comparison at its default settings.

edit ISO 800>8000

Photo AI is shaping up very nicely! It is definitely the king of noise reduction, sharpening and enlargement (1985 35mm color slide scan):

One thing I keep wanting though is automatic tonal/color correction. That is just as important as the other enhancements in a one-stop app.

There is an Enhancement option in PAI but I am not sure what it’s doing.

Here is an example of what I am talking about from another upcoming, unreleased app in Quick Edit mode:

Here is a portrait from Photo AI with nice enhancements but no color correction added or option to do so:

The other app:

1 Like

Further to this, it seems there is no way to disable autopilot for multiple images so it’s possible to process batches with the settings I prefer (resize 2x and enhance-natural only)… Can an option be provided to disable auto pilot for multiple image processing? Otherwise the “use these settings for all” option is somewhat misleading.

I have also been experiencing color shifts in image files processed as exports from lightroom using Edit a copy with lightroom adjustments. Its not a big deal since when it returns into lightroom the color is correct, but seeing other users/testers captures of the PAI interface, their colors appear correct.

2 mins. to install to Win11 Pro. Processor = AMD. Testing tiff image - Standalone version. Used AutoPilot.

A.P. detected “Severe Noise” - I don’t see any real noise. If stretching I’d accept Low Noise.

A.P. did Not Detect Subjects - Including the 2 well-fed men standing in the field… So, no masking to use if I wanted to selectively sharpen them. Which I would have done if could do incrementally.

2nd Image - ARW raw image - Standalone Win 11 Pro. AMD. Used AutoPilot, then Sharpened manually (to test the quasi masking feature from subject/object detect).

A.P. detected sculptural Subject on neutral background - detected sculpture & cast shadow as part of a single subject (which is fine as a detection - for masking, I’d want to be able to paint out the cast shadow to not be sharpened as part of the Subject - but as a 1st step toward masking I applaud :clap:t3:).

A.P. detected medium Noise - which is fine. But it got rid of most Luma noise & was less effective on the Color noise (see along edges and interior cutouts of the metal sculpture.

Test 3 - Final for tonite so others can rpt… - PAI .6 Ps 2022 Plugin test. ARW orig image opened (no adjustments) via ACR into Ps interface (I realize it won’t be processed as a Raw image…). Maintained processor = AMD (for spd). Autopilot initial followed by manual sharpening using Object Detected mask.

AutoPilot detected Subject/Object - mannequin in neutral museum setting. Was able to use detected area to further manually sharpen the mannequin.

No Face Detected - A.P. wasn’t faked out by the subtle features on the mannequin head. :wink:

Able to Save to Ps okay - saw change in sharpness quality from background layer to PAI layer. I was able to use Ps’ Object/Subject Detection Tool on the PAI output layer with no issues (eg., no harm done to the image by PAI processing that prevents further Ps noodling…)

Are you wanting to correct a color that was rendered incorrectly or enhance and ‘zhuzh’ up (a design-y term… :slight_smile: ) your images (ie, add more vibrance, saturation or other ‘pop’ adjustment); vs correct a color that in processing made a yellow appear orange, for ex.?