Autopilot seems to be hit or miss. This image was shot at ISO 2800 and definitely needs to be denoised. Autopilot did not “auto” activate noise reduction.
The image on the left was rendered with LR AI Denoise for comparison. I prefer PAI at this time due to its superior sharpening. However, I suspect LR will get an AI sharpener in a future release.
How is Autopilot trained to “see” noise? Is it possible for Autopilot to read the Metadata of a file and then apply the appropriate amount of Denoise?
Or, as a new feature, it would be nice to set up PAI to reduce “x” amount of noise for ISO “X.” I have Lightroom set up to do this with the Legacy noise reduction. If built into PAI, it would make batch processing a breeze.
honestly, I prefer your left image. There is more detail nothing is lost with Lightroom’s Ai denoise. On the other hand, I like the sharpness of the one on the right.
The strong model is really interesting on this picture. I also compared it with the motion blur which doesn’t look as good. But can do better on other type of image.
I love the “legacy models” approach here and in VEAI 3! This way the workflow we already have, won’t be interrupted, and everyone can find something they like.
Hello. I have been having an issue which I believe first appeared in this latest version (v1.3.1).
When running a .dng file through Photo AI from Lightroom in Windows, I use the File>Plug-In Extras>Process with Topaz Photo AI technique, which is recommended by Topaz (rather than right-clicking on the image and selecting Edit in>Topaz Photo AI). This has always worked for me just fine, up until a few days ago. When I do this now, Photo AI returns my processed .dng file to Lightroom with a significant blue color cast.
At first, I thought that I must have done an improper install, so I uninstalled Photo AI and reinstalled it. That did not resolve the issue.
On the left image below, you will see the .dng file before running it through Photo AI. On the right image below, you will see the .dng file after running it through Photo AI. As you can see, the color cast is substantial, and it is proving quite difficult to return it to its original cast.
Why can’t you use photo Ai by right clicking on the image when creating a Tiff?
For my part, it is what I do because the DNG have an annoying tendency to have some bugs with photo AI by going through “File>Plug-In Extras>Process with Topaz Photo AI”.
It’s good to have options to explore because different images get handled differently by the software, & not everyone wants their images processed to some robot “rule” or “standard” if they’re trying to be creative with their art (of any ilk…).
Thanks for your post, harald. Until I find a better solution, I will indeed right-click on the image and select Edit in>Topaz Photo AI. I prefer to use the File>Plug-In Extras>Process with Topaz Photo AI technique, though, because this is what Topaz recommends, as per this message:
The strange thing is that the File>Plug-In Extras>Process with Topaz Photo AI technique was working just fine for me up until a few days ago.
CDN? I don’t know what I know. But if your connection is between 2 and 5 Mb/s, it’s normal that you won’t get higher speeds when downloading the software models.
Of course, I adjusted the slider according to the effect I wanted. There are images where the motion blur works well and on others it is the strong. That’s why it’s good to have both .
You’re welcome, but although there is this message, the bug in question is known. And in the update logs, the Topaz team recommends right-clicking on the image and sending it to AI photo following the image extension formats in the dialog box that appears.
Saving as a TIFF we have 3 choices: no compression, LZW compression, or Zip compression. With no compression and Zip compression we can select 8-bit or 16-bit, but strangely for LZW compression we can only select 8-bit. Why?
Anthony mentioned the rationale at one point. I believe he provided some test data. But I, too, still wish there was the good old LZW 16-bit there used to be in my Topaz products and that all other photo processing software I use (it’s a lot from a mix of companies - CaptureOne & Lr not in my mix) still offers & supports. I prefer LZW compression b/c there is no hassle to launch that file type for me & I’d be jumping thru hoops with zip.
---------------- Here’s what I found from old threads:
"LZW in this case will often make images that are larger than not only ZIP but even uncompressed TIFFs. It’s standard practice to only use ZIP for high bit depth and LZW for low bit depth.
Have to agree, Eric.holbrook, with your comments about the denoise and sharpen tools. The upscale/enhance models seem to do a better job with noise removal and sharpening than the actual noise removal and sharpening tools themselves. It seems that as soon as I turn on denoise or sharpening the effect is undesirable. But I am still on version 1.2.3.
I have never been so apprehensive about updating a piece of software as I am with Photo AI. Every time a new release comes out I read the forums to see what issues have emerged with the new version, and then I hold back. Version 1.2.3 works ok for most of my applications (as long as I don’t use denoise or sharpen).
I am optimistic though: the Topaz engineers are listening, and they are tweaking away, and one day they will get it right (though it is a worry when I hear people on the forums say that earlier versions of Gigapixel yielded better results than the present version. Now that’s a mystery . . .)
.