It’s another weekly update, this time with polish and bug fixes, as well as an improved model for detecting blur in images!
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.
Released December 29, 2022
Changes since 1.1.3:
- Updated blur detection model
- Added hotkey to switch masking modes (x swaps between add/sub)
- Checking for updates button is faster, and has a manual refresh button
- Improve main menu organization
- Menu items now show shortcut keys on Windows
- Removed autopilot green dot when the autopilot suggested model is NOT selected
- Apply Current Settings to All now respects Linear RAWs
- Fixed Mac not closing images correctly
- Fixed split view being offset at 100% display scale
- Fixed plugin crashing when using older hosts (CS6, Corel PSP)
- Fixed some images using the wrong camera profiles (e.g., Canon EOS R5)
- Fixed old preview image popping in while processing a new image in batch mode
Image shift is still very apparent when upscaling 2x and moving preview slider.
If you pan all the way down to bottom or top left corners of image, the shifting stops.
Great ^^, just installed it and will check.
Can you check on the following in the future? Sometimes when i add a few pictures, some faces are not detected, but I only notice if I manually check each one of them, I can manually add them by selecting the face. Could you make an option to maybe after conversion, list files for which faces were not detected? Or maybe a pre-check before converting? Anything that assist without having to go images 1 by 1 to confirm detection will work.
I may be embarrassing myself because I’ve missed this, but is there an option to not automatically refresh an image preview while you’re changing settings? The automatic preview update slows things down considerably, and I’d love to be able to set everything just as I want it then manually refresh the update when I’m ready, like you can do with other Topaz software.
It would also be very nice to be able to save customized settings to apply to other images. Again, if this feature exsists and I’m missing it, forgive me.
Win 11 Pro desktop PC. Processor = Auto.
Installed & launched okay (in Standalone & Ps 2023 Plugin). Able to save from Standalone. Able to save back to Ps layer from Plugin.
A.P. didn’t sense need for Sharpening in 1st image test. But did denoise color noise:
I tried Masking the man in the scene. AI Masking Brush continues to color outside the lines in some cases & doesn’t reach to desired edges smoothly (stays some margin in and is jagged). I tried diff. application methods…painting vs clicking + enlarging (zooming in on) figure vs working zoomed out to see if there was a method that would let me accurately mask. Unfortunately not. But I was able to use the “X” hotkey to alternate b/wn Add/Subt mask overlay (Thx!).
Once masked as well as the AI Brush would allow (it was clear I couldn’t get the mask how I wanted) I activated Sharpen Subj Only. The preview processing still seemed to sharpen the mid-ground in front of the man too - adding more details to the ground.
Also a few times when I hit “Clear” for the Masking strokes, nothing happened and the overlay remained. It was inconsistent bhvr in that regard…
I was able to take the pre-processed, saved output image into Ps and post-process it some more. It didn’t become an unrecognizable, unchangeable file format:
BTW, including Shortcuts in Menu items is useful!
Thank you for fixing the plug in crashes with PaintShop Pro 2023. The Photo AI plug In is working again. Love the Topaz software you guys developed.
I can confirm that Photo AI is once again working with PSP2023.
At first I thought the split screen was not working as the slider bar was over a totally white section of the image and since it is also white it did not shop up.
Hi, can someone please advise if it is possible to open an image from FastStone image viewer (or opened from another app) to TP but once rendered save to another or open in another app direct from TP? Thankyou.
Confirming that v1.1.4 works as a plug-in with On1 PR2023. However, TPAI is still not much use to me until I have the use of a more refined masking brush. The fact that the devs never acknowledge this as an issue indicates to me that they are not interested in listening to our concerns!
In V 1.14 still not fixed:
It ignores the lens profile integrated in the CR2 raw file of my Canon Powershot G7X Mark II, even though “Enable Lens Correction” is activated!
And also in V1.14 still NO thumbnails are displayed for Canon CR2 and CR3 files in Stand Alone mode.
(Windows 10, latest Update, RTX2060)
On Mac Studio M1 Max: Apply settings to all for more than 500 photo’s, the UI freezes when clicking the confirmation popup.
The popup also stays visible when you switch to other applications
When using the edit in option in Lightroom I loose al the meta data.
I think that a simple brush with good edge detection would be better for masking. AI often gets it wrong and needs correction and so far this is not possible in PAI.
Have you “Liked” my proposal for masking brush features yet (if you agree with them) - on the PAI Product Roadmap discussion thread?
Last I looked it was over 11 likes… to let devs know there’s a wishful (for specific features) community
I think of the number of Likes there as comparable to dog years - each ‘Like’ (Heart symbol tap) represents a large multiple of users/customers who don’t even know about these forums or how to maneuver through them. To my way of thinking, each of our Likes are omsbudmen for at least 100 other users.
Fotomaker I think perhaps the time has come for compromise. I understand the stance that an AI company should have AI solutions…however there is a point that idealism needs to be tempered with practical function. The AI selection brush in TPAI has swung from a waste of time , to something that is almost correct in its selection much of the time. This to me beats all the tedious brush strokes of a formal brush and as such is successful. The time has come to now realise that its potential is reached for now yet our practical requirement for fine selection is not resolved. This to me is the real test of pragmatism, the developer now needs to satisfy the client and bend his ego enough to allow an 90-% AI solution for the heavy lifting , finished and complimented by traditional finishing. The perfect marriage where each technology brings its unique advantages to a fast and functional solution.
May I add comment on the CS6 plugin and its re found function… thank you !
Seems to be working well as a plugin in Affinity Photo. I’m wondering about lens corrections. Photo AI would almost never be my last processing step, so is there any reason to do lens corrections here, vs later?
There is just no current way to accurately mask a subject or object with the AI Brush. Unless, there are distinct tonal boundaries between said subject/object(s) & their surroundings. A lot of photographs do not possess that. Especially if PAI is intended as a preprocessor; which was my sense it is supposed to be, given its functions.
I, personally, don’t like the concept of relinquishing creatvity to a (as I like to call it) ‘robot’ (a computing engine that ppl attempt to train but it uses math to interpret that). To me, that’s orthogonal to making art.
That said, I’m not opposed to using a Smart Brush; if the brush behaves smartly & saves me time in achieving my vision.
To me, the needed, efficient smartness I would value is a Smart Brush that I could use as a choice or on demand to fill a mask accurately to edges my hands might not be able to paint - ie, to provide a very accurate Edge Aware functionality selectively & only where desired.
Unfortunately, the current tool is doing anything but that (per my description of what I’ve observed above)
If there could be a genuine integration of manual, controllable brush (with other attributes I noted in my wishlist in the Product Roadmap thread) & further refinement, for accuracy, of the nascent AI (Smart Fill) Brush, that would be a best of both worlds & probably a highly competitive product offering. IMO…
There are still issues with the “recommended” method for Lightroom…
- There is a noticeable pixel shift
- Even after copying all the settings (including all masks) from the original, there is a noticeable color shift and a difference in e.g. the background.
Here’s what I did:
- Took an already edited but uncropped (6000x4000) photo, selected File, Plugin Extras > Process with Topaz Photo AI, selected “Edit a Copy”
- The resolution of the copy in Photo AI was showing as 6024x4022
- Done the Noise Reduction and Sharpening using Autopilot, saved the copy
- Back in Lightroom the 2 images side-by-side in Group View looked like this - for some reason the copy’s resolution is showing as 6000x4000 again:
- Loupe view:
- Develop module:
- After copying over ALL the settings from the original, there are visible differences between the 2 images:
- Also the lens information in the Metadata is messed up:
…which still leaves me with having to use the “old” method (process a TIF with LR adjustments).
Always going to be the best way as you retain the RAW conversion of the App you use and like. No 2 RAW converters will ever be the same.
For Topaz products I never use a RAW file as input except for Beta testing and, even then, don’t output DNG.
(As a side note, I noticed it happens more with close up pictures but not as often with smaller faces on the frames)