It’s likely like you suspect not supported RAW file yet.
2 Likes
B/c the updated ‘Gigapixel’ features - as a plugin to Ps - are better than Ps’ scaling features. IMO. That’s why it’s critical. Also (usually, but not for every image) I prefer the Topaz denoising to Ps’. I very infrequently ‘shoot’ group people shots & ditto post-pandemic, with increased street crime, street photographs that include smaller faces. But PAI is better than anything I’ve used in Ps to restore or sharpen up faces.
With my entirely Ps workflow, I rely on high caliber Topaz (& other 3rd party) plugins to be efficient. Both traditional Topaz creative ones & current functional ones.
BTW, I believe this will become even more important if/when Ps releases to the masses some of the Firefly features they’ve been mass beta testing since earlier this year. That will gen low res content that needs ‘cleaning’ & scaling.
They use Lensfun. And seem to update the version of it to support newer gear on a regular basis.
Is your gadget on the Lensfun supported list. You can check it via a Google search of Lensfun.
1 Like
crash logs for face detection failures. Tried using both GPU (nvidia 3060) as well as CPU models just in case. offending image is also attached here. 33 faces granted, but I’ve done even more successfully. =) Let me know if I can provide anything else! I have even more log files on this as I have attempted different file formats, models, AI processors, etc. I can only upload 2 files though.
2023-09-12-12-26-7.tzlog (11.6 KB)
2023-09-12-10-05-40.tzlog (182.1 KB)
2023-09-12-10-08-1.tzlog (1.7 MB)
This looks like it borrowing color from the squirrel and adding it to the grass.
2 Likes
I’m also still seeing this huge RAW color shift issue in the latest 2.0 version. When will this finally be fixed?
2 Likes
Its NVIDIA not TPAI thats the problem
I agree with devenpsc - TPAI has outgrown it’s original focus - denoise, sharpen and upscaling - and in doing so has lost the niche, lost the quality, lost users, and IMHO lost focus on what Topaz does well.
I liked the idea of TPAI having all three products in a single piece of software. I’ve stuck with it and I “hope” TPAI is getting there with that core focus.
I dont like the add-ons that will always be done better by products that focus on RAW editing, and in my view not what Topaz is for.
I’d be extremely happy if there was a refocus on the 3 core products in TPAI.
Not sure if the engineers at Topaz are aware - the original Denoise and Sharpen are still producing better, reliable and repeatable outputs that TPAI, even in V2.0 cant do!!!
I guess, I am, along with many users, frustrated that a great original product has gone through so much develoment and not really shown any significant improvement.
10 Likes
And that is the winning statement!
5 Likes
Which is honestly their greatest weakness if they are. Relaying way too much on open-source community to add support and to fix bugs. The glass list is actually pretty damn small and isn’t being updated regularly from my check. Explains why having Lightroom export to TIFF instead of DNG works so, so much better for me for their stuff. Especially the annoying color shift when working with RAW files that happens.
Assuming that is actually accurate they are relaying on that.
There’s much to improve upon the original core features of Photo AI. I hope Topaz developers focus on stability and enhancement of existing features for a better user experience, rather than experimental new features. Especially considering the questionable usefulness and accuracy of colour balancing.
I know this doesn’t help but calling PAI v2.0 from within Affinity Photo 1.0 works fine for me. Possibly re-install PAI.
1 Like
Using a Sony ARW raw file there is no color shift if Balance color is off. Below is what happens with Balance turned on. You didn’t state the condition with Balance color and what raw file was used.
I’ve been away from my desk for 10 days and kinda glad I didn’t download v2; although I am glad it fills a need for some.
Many people using an updated or trial version of PAI 2 will be disappointed and will look elsewhere. Cognitive Dissonance theory explains a lot of consumer behavior and can bite a lot of vendors who aren’t keeping that in mind.
There are a number of professional products that perform the newest features of PAI better and faster. AND there are more amateur oriented products that can do them better and faster, too; even free ones.
BUT, how often do you see a rollback of new features by a developer?
1 Like
I’m still getting tiling artifacts with the new release. They don’t seem to be as strong as before but they still exist.
Topaz Photo AI ver 2.0.0
If I had a magic fairy wand and could make Photo AI V 2.0 into what I would have wanted, here it is.
And, I would love to throw money at the developers to have these core type features of my fairy wand wishes actually work at 100%.
( Btw, I do landscape nature work, not facial enlargements, so I won’t comment on face stuff.)
Being able to blow up photos even larger than 6 X and have those blow ups be true and accurate to the photos with lovely detail, as if I had shot whatever older image with a $10K lens let’s say.
In other words, no hallucinated weird leaves and wrong sized leaves/needles/grass/moss, and so on.
Also, no furry rocks, furry water, furry pine needles and so on.
No giant square hunks of complete blurry areas right next to perfectly sharp areas, despite no depth changes.
I would love to have the software much better trained to understand in the first place, what a rock is, or tree/log, water, etc, and where one ends and another begins. Also, a much nicer/detailed and realistic blow up of water, or water falls.
I would love to have boulders that blow up with the fine grain of boulders and not have the green/yellow, etc., moss on the boulder turned to big leaves. Also, not have the rock area itself look like large clumps of clay stuck together.
Less baffling controls/tools that often don’t make sense in the context they are portrayed as, nor when you try to mix them.
For instance, the new sharpen V 2, the more I turn it up, the more PAI V2 hallucinates new strange and wrong sized leaves/grass/moss? I thought it was a sharpen tool, not an add more leaf, etc., models tool.
I’ve dinked around with PAI for countless hours and I still am not certain at times what slider is doing what in conjunction with the other sliders. I don’t have that problem in any other photo editing software. It does not have a logical flow, at least not for me.
I would like to not have to add blur to already sharp photos to have the software decide it can only add the models for grass, leaves, rocks, etc., after I add blur.
I would love to have a working, professional quality level of batch processing with nuanced controls and that can handle thousands, or at least hundreds of old photos accurately and without crashing.
I would love to have presets I can make and save that are specific for me and not tied to auto pilot in any way. And I would have a complete turn off feature for anything auto pilot, which I thought V2 was supposed to have, but does not near as I can tell. I’ve shut off everything I can find, and there’s still auto pilot stuff going on.
I would love to have a normal sharpening ability, like a tiny radius setting like PS has so I can finely adjust the thickness of the sharp edges, such that blown up blades of grass don’t look like green twigs, etc., the way it is now.
There’s more, but you get the idea. As I said before in another comment, I don’t need Photo AI to be Photoshop, or Capture One, Lightroom, etc. I just want to blow up beautiful photos with more detail. It is WHY I bought the software.
You gave me a small taste of what could be in “parts” of some of my photos, but then that gift gets taken away in other parts of the same photos, all photos, without exception. All versions of the software back through Gigapixel, without exception, give, and then take away, the thrill of what could be.
Please continue work on the core reason most of us bought the software, to blow up photos and add fine detail that wasn’t there. Then, keep going, making those core features better and more awesome. Anything else feature wise doesn’t make any sense to me for why I spent my money here.
There’s my fairy wand wish list.
I wanted to add. Imo, the key to being strong in the market, is honing in on those one or two things that no other company is doing, or doing well, and drill down that specialized development hole harder and faster than anyone else perpetually.
The more you diversify beyond what attracted users in the first place, the more you spread yourself thin, water down what is unique about your company and lose that specialization or obvious advances over your competition, and thus, lose your user base.
11 Likes
LibRaw.
Lensfun is for lens correction.
I would love to have a working, professional quality level of batch processing with nuanced controls and that can handle thousands, or at least hundreds of old photos accurately and without crashing.
I would love to have presets I can make and save that are specific for me and not tied to auto pilot in any way.
Those two just about sums it up for me as well. f those two worked I’d be a happy camper.
I would on the other hand also like (in addition to being able to turn it off completely when needed) to be able to trust the Autopilot enough to just “fire and forget” on larger jobs… I mean - shouldn’t the combination of AI and Autopilot by definition mean just that? Load a job, click process, and trust AI to do it’s magic?
2 Likes
I agree with devenpsc.
That’s always the great danger of building a product. Starting small with core functionality, which has great potential, adding more and more features, until one day you have a mega-complex software, that only works half way. You might then find yourself in a situation, where the competitors kept focusing on their core functionality enhancing it over the course of time - in a way you now cannot make up for it.
The way it goes right now, with all the added new features lightning, color balance, and so on; it won’t be far that there will be a healing brush, a clone brush - and at one point you will have a fully fledged integrated pixel editor. 
But the core functions will have suffered massively (as I would assume that you will not have stocked up the team exponentially).
There’s already a ton of proven products out there that do all of that editing and also artistic enhancements (Adobe PS/LightRoom, CaptureOne, Affinity, Luminar AI and so on).
As devenpsc stated the blurry patches issue and others are still unsolved.
Working with large number of images (batch processing) is still not usable in any way, which worked great in the old DeNoise/Sharpen/Gigapixel product line).
CLI is still very basic and cannot be parameterized, see here
Please focus on the core functionality, so that is stays an AI photo enhancement application. With the focus of making images look as good as they can - while keeping a natural look to the people in them (avoiding a plastic Barbie look 
)
And allowing users to work with a large large number of batches (UI and CLI).
That’s all we really need, at least from my perspective.
Plus enhanced quality control (as suggested in this idea, feel free to vote for it)
3 Likes
Thanks but, as I said above, applying the fix suggested by Krbo sorted the issue & I can call PhotoAI 2 from both release & beta versions of AP 2.
