PAI vs. the just-released ON1 Photo RAW 2023.5:
Resize test #1 (4X). ON1 at left, Topaz at right. Source photo was an AI-generated image I did using DiffusionBee and PS beta. You decide 
Original full-frame source render @ 100%:
Resize test #2 (4X, with Face Recovery). Topaz at left, ON1 at right. ON1 is doing some poor āMr. Potato Headā workā¦ You can see the original image at far left.
Another, same test and orientation:
Thanks!
In an earlier version - I believe it was version 1.3.4 the last time I saw it - the size (X) changed when I resized the photo in pixels:
That doesnāt happen anymore.
Can this feature come back?
It is either an issue with the type of images you are processing, e.g. RAW vs TIFF/PNG/JPEG or processed in another app where Noise Reduction has been applied, or the GPU driver. This is a ISO25600 image that processes fine:
Letās say I am working with a LARGE batch of images, the Autopilot it selects, isnāt always the best/optimal for the image. Yes, you āApply Current Settings to allā and it will apply the enhance type to all, BUT it will copy the sliders that image you āapplied allā had which may not be optimal for others.
TLDR; Is it possible or Would it be possible to implement the ability to, while batch processing, preset the enhance type for batch images, while allowing the sliders to Autopilot for each.
Lightroom, no noise reduction applied, TIFF.
It shouldnāt be that hard for PAI to read the metadata to see the amount of noise.
Or to āseeā the noise regardless of the file type.
Also, Iām not referring to PAI being unable to remove noise. Autopilot does not āautoā turn on noise reduction.
Excellent! I also struggled with FFT and with not very good results.
I own both Topaz PAI and ON1 Photoraw 2023. ON1 is new to face recovery and is a little behind Topaz. Look at the personās eyes in the last photo. I expect ON1 to catch up on face recovery soon though. I still prefer Topaz and it can be used as a plugin to ON1 Photoraw but canāt be upscaled.
Thatās right, and if we didnāt have PAI to compare it to, ON1 would seem a good choice ;-).
Of course Photo RAW does a lot of artsy things PAI doesnāt do (Sky Swap, etc.) but in terms of denoising, sharpening, face enhancement and enlarging, PAI is still best in class.
Iām an automotive action photographer and have some wish list items and issues to report:
Wishlist: Add a motion blur correction option to the sharpening function within Photo AI (or add a denoise all image with local sharpening in Sharpen AI).
Issues: Graphics with any amount of motion blur get distorted with sharpening in Photo AI. It gets 10x worse when combining noise reduction and sharpening in the same image. When I refer to graphics Iām referring to items that are very important to get correct on racing images such as sponsor names / logos, driver names, car numbers, etc. Recommend working with images of this type to see if program can be improved in its handling of moving objects in the image. Focus on what happens to text in images when sharpening and applying noise reduction together.
Huge Kudos: The primary subject auto selection does an amazing job. Cleaning up and refining is better in Sharpen AI than Photo AI, but both are pretty impressive. Noise reduction has been working wonderfully.
Processing high resolution JPG images with Adobe Photoshop 24.5.0 on a Windows PC. Photo AI has been very stable on my system.
Hey Mike, could you post an example or link one?
I would like to see how it does look like.
I need to know the type of lens and aperture, ISO and shutter speed also.
Thomas,
Iāve attached two images, before and after applying default sharpening and noise reduction together using Topaz Photo AI v1.3.6 with the Normal V2 (beta) noise setting. Zoom in on the carās license plate and the text graphics below the driverās door. Most of the sharpening works well, but the text is not great.
As for camera gear and settings, those will be changing constantly during an event based on lighting conditions, car speeds, distance to target, car direction, and intended effect. Sample image data:
Canon 90D
Canon EF70-200mm +2x converter
Focal length 140mm
f/10.0, 1/160th second, ISO 100
thx.
You probably want to keep the blur caused by the dragger, you just want the writing to be sharp and legible again?
Otherwise, A.I. denoising has changed the way we photograph.
We always underexpose by two stops (2x 1/3 stop increments, i.e. just before -1 ev) and use Auto ISO with aperture priority.
My hopes are that the sharpening step can be local to the selected item (car), and that the denoise step would be for the full image without enhancing details (or minimal enhancing). With racing photos you need to maintain the background blur and wheel rotation blur, but keep the main subject in focus, clear and sharp. When working with my selected main subject I will remove the wheels from the selection prior to applying the mask, to maintain the wheel rotation blur. I read elsewhere that Topaz was working on a text recognition feature which Iām hoping can be used in each of the Photo AI, Denoise AI and Sharpen AI filters.
Iām mostly with you, Mike! But - perhaps selfishly - Iād also like the ability to Mask and select where denoising is applied too.
Sorta like the āclassicā Topaz products had an Apply button where you could select some function in the program to operate on a specific area of an image. Then Apply that to essentially create a new base image. Then move on to the next step of processing - perhaps adding selective Sharpening. But I guess thatās my Wishlist Item for the product Roadmap vs comment on this releaseā¦ It didnāt add layers (not like Studio) but essentially it functioned like adding a layer then flattening the layers to continue - if you use Ps at all for the analogy.
