Windows 11. Camera is Sony a7IV. I save images as lossless compressed AWR files on SD. I convert to DNG on import into Lightroom Classic (most current version). Using recommended File-PlugInExtras-Process with Topaz Ai method to invoke Topaz Photo AI using DNG image. Resulting image has artifacts horizontally at the bottom and vertically at bottom right of image - visible in DNG returned to LrC and in Topaz Photo AI.
Only occurs with DNG original. Tested with older Sony raw format from A6000 and A7III cameras. Topaz AI worked flawlessly and returned clean DNG.
Problem does not occur to the same image in original AWR format I retrieved from SD card.
Suspected bug is in the conversion of Sony AWR images to DNG format during import in Lightroom Classic. Artifacts added in reconverting image to DNG.
Temporary fix is to delete original DNG from LrC and reimport raw files without conversion to DNG. But I have >12k DNGs.
Like that the ability to edit LR HDR files has been added as when I want something quick and not up for changing lenses I tend to go the route of pulling out my phone and using the LR Mobile camera app.
But I have noticed that running these files through Photo AI brings it back much darker even with only slight differences in the settings once it is reloaded back into the LR catalog
I have a few thousand images to batch process and using the GUI it is almost impossible. The more images you add to the queue the slower it gets to the point it is unrealistic to process anything.
I did a test - I added different number of images (720 x 480) jpeg to the queue and in the settings only enabled Denoise to all images (all other settings disabled), saved output as JPEG Quality 100.
Closed all images between runs
To process 10 images it took about 5 sec or 0.5 sec / image
To process 100 it took 73 sec or 0.73 sec / image
With 1000 images it took 2897 sec or 2.897 sec / image
As you see, the bigger the batch size, the longer it takes to process individual images. There is definitely something wrong with the application code here. It is probably doing something inefficient involving all files in the batch such as enumerating all files every single time you process a single file or something like that.
Topaz Photo AI [v1.2.4] on [Windows 11 x64 with 32 GB RAM]
GPU processing using NVIDIA RTX 2060 Super
EDIT: I compared with Denoise AI v3.7.2 and it absolutely kills Photo AI in terms of speed. I added the same 1000 test images to the queue and it zipped right through them completing them all in 5 min 21 sec = 321 sec = 0.321 sec / image. Wow. There was no noticeable difference in speed per image when changing the number of images in the queue.
If I load a number of images and some of them are to be upsized (by auto-upscaling to a specified minimum width in Configure Autopilot) then the descriptions in the save dialog correctly state that the smaller images will be upsized but most of the saved images have not been upsized. (I think it may depend on having previewed the image: if I haven’t explicitly visited each image to preview them then none are upsized but if I preview them all one by one that seems to overcome the problem.)
Apologies if I’ve missed it somewhere, but I can’t find any way to save just a selected subset of images: I generally load a number of images and initially save them all with autopilot settings. Often there are a few images that I’d like to manually tweak the settings for before re-saving just those. But Clicking Save always seems to result in saving every image in the batch. It would be very useful to be able to save a single image or a manually selected subset of currently open images.
(Otherwise I am very pleased with how Photo AI has progressed: initially I continued to prefer the older Topaz programs for their more detailed control capabilities and because of a lack of some features that I wanted. But for a long time now Photo AI has met most my requirements and certainly makes things easier.)
[Please describe your issue here. Try to include logs (Help → Logging) and add screenshots if necessary.]
I have a fast laptop running Windows 10. I tried to install the newest version of Photo Ai and it would not install. When I went back to try to use the previous version which had successfully installed a month ago, it would no longer work. I have tried to download and install 3 previous versions of Photo Ai and I keep getting an error message that an install application is still running and it cannot install the program. I am now locked out of my software.
I crop this picture and now i see the pixels from my camera, the Sony A1?
or is it something else? They are all squares. It only happens with Topaz Photo AI in .arw (sony), not in .DNG… Would you look at it? Keep up the good work.
I compared view inside Sharpen AI and Photo AI
I don’t understand the issue with the zoom factor as you can see on this screen shot
Sharpen AI and Photo AI are at zoom 100% . The size of Photo AI display is for me two times too large !
samething with denoise AI also two times too large
This is something I complained long time ago for Denoise AI but never checked it is good with Sharpen AI So it will be good if you can be good and coherent between your applications
Concerning the Basic brush with PAI 1.12.4 I agree this is better than nothing but what I expect is a cursor like the one in Sharpen AI with edge detection I found much more usefull . I Hope it will came
It’s not. Gigapixel is based on the same software as Denoise and Sharpen so all three appear to handle files the same way. Photo AI is very different the way it handles the queue of files.
I am trying to use photo ai from lightroom classic in raw mode to enhance especially lowlight and high iso pictures.
For portability and when I’m out with my little son, I use a sony zv1. The one inch sensor pictures get improved greatly by photo ai. But i cannot use them because of lens correction. All my pictures are distorted and uncorrected when they come back to lightroom.
It does not matter whether lens correction is enabled or disabled in photo ai. It is always wrong the same way. I hoped disabling lens correction in topaz would allow to enable the one in lightroom, but no luck. Is there a chance to fix this? Best would be to be able to use lightrooms lens correction on the topaz dng. Is this possible? Could one inject the lens correction data to the output dng?
Hello Don,
thanks for your reply. I tried to convert my .arw raw to dng via lightroom and invoke the plugin. I had no luck - the same distortion happens.
I know that lens correction data is embedded in rx100 and zv-1 raw files, but I also know that rx100 is supported by lensfun also, which is used under the hood of photo ai for lens correction. So I tried an RX100 raw with the same result - lens correction is not applied by photo ai - no matter whether it is toggled on or off in the menu, processed pictures look the same.
So I’ve tried Rawtherapee - which also uses lensfun db - with the RX100 input image. Success, Lens distortion is handled well if switched on and camera is detected full auto. ZV-1 is not supported currently in lensfun, but I changed that locally and added it to the lensfun xml with rx100M3 lens data. That also worked and now ZV1 is automatically detected and corrected in Rawtherapee. The same modification does not work for photo ai.
So to sum up: Neither RX100M3 nor ZV-1 get corrected in photo ai, though RX100 is supported in the shipped lensfun db.
I did a little bit of further testing and I am quite sure that this is a bug - although I don’t know how long it is existing already.
It seems that lensfun is not called when photo ai is invoked via the lightroom plugin (module) for the raw workflow and so no lens correction is applied. This is visible in the logs, also. When used in standalone mode, lensfun is used and the correct profile for the sony rx100 is detected and applied. In my case, the ZV-1 is also detected and corrected because i added it to lensfun. However - problem is the same for RX100M3 without lensfun modification.
I’ve attached both logs - standalone and plugin mode where you can see the difference.
On a side note: Though lens correction is working okay in standalone mode, when you toggle lens correction On=>Off=>On in the menu, the preview correction is On=>Off=>Off. However, export is correct everytime correction was set to on in the menu.
So is the image passed to Photo AI from Lightroom via the Plugin the original RAW or DNG, because if it is the DNG lens corrections will not be applied. (I think)
Today in the office on a 2017 SSD iMac running Ventura latest I am experiencing the same thing I went through on the M2 Mac Mini at home: The app’s interface in standalone mode can’t be moved around the screen (it jitters and flickers).
Even worse, I cannot load an image with the Browse option (nothing happens). Same with dragging from Finder (whether TIFF or JPEG).
When used as a Photoshop 2023 plugin, Photo AI won’t load the photo (only the Drop/Browse window is presented).
Also, how do I save changes to nothing-?
I fought with PAI as a plugin some more and had to resize the window before the photo displayed. Otherwise it was No Filter Selected (or similar) and a blank canvas.
