This is exactly how I see it.
2 Likes
Yes, 33 votes is not too shabby - but youād need a counter-vote to really see how users think.
Plus you would not get the opinion of those that just use the software and donāt post here (after all there are 800 downloads alone for the 2.0.2 here in the respective thread which makes 30+ votes less impressive).
Myself, Iām quite undecided/ambivalent here .
Of course Iād like the original main features of the software be stable/bug free and with the best possible quality - but I donāt have anything against new features,either. And especially object removal is a thing where AI could really help IMO.
1 Like
I just donāt like re-invented wheels, thatās allā¦ I see it as a waste of resources.
But everyone has to decide for themselves.
Iāve done all that I can, trying to focus the problems and needs of the userās Iāve interacted with so far into this thread/vote. We will see, what comes of it. I am keeping my fingers crossed.
1 Like
33 votes x $100 annual subscription = 1 programmerās week of work / year.
Me Iām not really against getting new functions and other types of improvements, but when every new version - for some reason - seem to bring new references to version 1.3.x or 1.5.x as being much better in their corresponding forum threads, and I keep on reading complaints about regressions or old errors not getting fixed itās not really that far off to see that something aināt right with regards to the priorities made by the devs.
And Iām sorry, but the returning comment about ānothing is really forcing you to updateā¦ā
I paid for a years of updates, not for a year of regressions and new fancy stuff that only half work.
8 Likes
In Europe 3.700 is a programmerās monthly salary (as a permanent employee, not being a freelancer). In America it will probably be higher, but the costs of living there are also higher in comparison.
But we canāt use the number of votes for calculations anyhow; itās just an indicator, like an election forecast. Most users are not as active and involved as we are in this forum, so this is just a fraction which would need to be extrapolated.
And we canāt use the number of downloads either, as every single click counts (I myself have downloaded the same version 3 or 4 times already, which triggers the counter again).
But at least the tendency says something: This post is only one week old and has already gotten 37 votes, and is still going up. Compared to that, the second highest vote (iPad), has only gotten 20 votes; and it took over seven years to get these. All Iām saying is that it underlines the pains of the active and involved users - not more.
We managed to point out our problems and have been heard. Thatās all I wanted.
The rest is up to Topaz Labs. Itās like they say: Anything goes! 
Totally agree and have voted accordingly
Jim Greenfield
I fully do understand your point, BUT:
-
If it was like āin the good old timesā with the separate Denoise, Sharpen,ā¦ products we now would have an own RetouchAI app.
The effect then would be (and also always was even in that praised past) the same: that some time of development would go into the new app even though the original apps keep focused on their primary target. -
And imagine if e.g. Adobe Photoshop never had introduced any new features where would the app be now? Would it even still exist?
Of course this doesnāt mean that fixing severe bugs shouldnāt have priority. But itās not 100 or 0 thereās also like 30ā¦
1 Like
Iām not saying no more new features - ever.
We all agreed on new features that make pictures more look realistic are always welcome.
I only said - think twice before introducing and following through a mega feature like object removal - that can become an endless open wound (and look at the results so far in the beta chats compared to open source and commercial solutions already out there).
Itās no problem if just 20% of the staff work on that and the rest works on the core features and bugs. No doubt. But Iām suspecting itās more along the lines of 80% new feat / rest bugfixes. The Pareto principal can usually be appliedā¦
Never the less, to stop future discussions on this with you - I will hereby say - that you are right, you won 
. Iām not here to argue semantics. The only goal I had was to make Topaz Labs aware of our pain points. There are now fully and undeniably aware. And it is now in their hands.
I will now also have to cut back on the time, that I invest in this. I do have a day job as a senior software engineer, tooā¦ I will look back here from time to time, if there is actually some progress on the matter. And then at the time the subscription reaches its end, I will decide for myself if it was enough effort in the right direction or not. As will everyone else here have to do at the appointed time.
Over and out for now
1 Like
Itās true that we wonāt get a representative sample of users just from threads here, but you could make a counter-thread and see where that goes.
I suspect that thereās a substantial āhiddenā demand for image editing features (a āPhotoshop/Affinity liteā as @PWC mentioned above) among many users who arenāt active here, which is one reason why Topaz is going in that direction.
I think this point is extra important in the age of rapid AI development. Generally, it shouldnāt be the case that people look to some old version (including Gigapixel/DeNoise/Sharpen) as better than the current version of Photo AI, except for nostalgia or some corner cases.
The direction and feature overlap with other apps is my overarching concern at this time. The broad capability of the object removal feature is a strong sign that Photo AI is going the image editor route.
Even a focused Photo AI (stage 3 or even 4 in my earlier post) can benefit from object removal if it is used to create 3D photos (a pair of photos) from a single photo. Presumably, the parallax difference between the two photos requires some objects (or parts of objects) to be removed while other areas need to be filled in by AI that guesses what is behind some foreground object.
Disclaimer: Iām not suggesting that Topaz work on adding an AI-based 3D photo creation feature at this time. Itās just an example to show that object removal can still fit in with Gigapixelās focus on upscaling, since 3D photos and depth maps can be thought of as āupscalingā the ādepth resolutionā of a photo.
I believe that this thread is a tipping point for many prospective Photo AI users who originally downloaded this app for its unique features. However, Topaz Photo AI is trying to add new features to enter a completely saturated market and compete with other top players. Unfortunately, these new features are coming at the expense of the development of the appās core features, which gave it the identity it has now.
I believe Topaz should recognize that itās unrealistic to keep being a good photo AI while competing with top-of-the-line photo editors in the market (like Adobe Photoshop). The current trajectory will contribute to brand dilution while failing to compete with existing apps in the market that are superior and better recognized.
Thus, I hope Topaz recognizes its priorities for Photo AI, as the appās core userbase is starting to feel abandoned. People who share about Photo AI on social media highlight its core features, not photo editing options. I am not asking the Photo AI team to cease the development of new features, but perhaps it would be ideal to scale back and focus on whatās important.
6 Likes
Perhaps we should close the āIdeasā Category entirely for a while, so they wonāt be tempted by our user requests any longer? Take a look over the last year, and youāll see where a lot of this evolving scope of work is coming from. For example, Automated Horizon Leveling recently popped up. Seems simple enough to go along with the Crop and Rotate everyone seems to need as part of the core features. Alternatively, perhaps we need two voting buttons, Yes and No, so all these good but distracting ideas can be voted down.
1 Like
When TPAI came out I was struck by the possibility of having all the features of previous AI products contained in a single app.
For me, PhotoAI seemed to be a plugin for the most well-known image editors (Adobe Photoshop and Lightroom, CaptureOne and many others) that allowed an improvement in terms of
- denoise
- sharpening
- face recovery
- upscaling
This was intended to improve the quality of images where traditional editors were not skilled enough: so as a plugin it added power to them.
As time went by, however, I saw that TPAI wanted to become more and more an additional image editor and here I believe is its current weakness: as good as the developers are, I believe that it lacks the critical mass to be able to compete with for example Adobe: it would better that Topaz continued to invest only with the functions on which it had distinguished itself with the individual apps, niche but notable, and abandoned the idea of competing with Adobe and others.
I hope to be able to use TPAI only as a plugin to improve my images after having processed them with market-leading, more functionally complete and constantly evolving editing apps.[Please describe your idea and how it would be useful to you in your workflow.]
7 Likes
I am increasingly convinced that Topaz has decided with PhotoAI to integrate the products it had (Gigapixel, Denoise, Sharpen) and in which it excelled, and to always add new editing features to them to compete with the other image editors that have been on the market for years .
I think that a worse move could not have been made: the limitations complained about by many users, for example on face recovery, are not overcome and attention is paid to features already present in all existing image editors. So past users donāt see the features that excited them improve and instead see things they already have available in other apps. The RAW converter itself certainly cannot compete with those of Adobe, CaptureOne or others. Wouldnāt it be better if PhotoAI limited itself to applying its original features directly to TIFF or JPG, for example, and limited itself to being a valuable plugin for existing publishers?
For months now, every week a new version has been released with even heavy regressions compared to the previous version: therefore, users, instead of receiving a complete and tested product, must themselves, every week, be beta testers who provide feedback to the developers, therefore paying a product for testing.
To conclude, I donāt think I want to spend $199, which already seems like a considerable amount, for an additional image editor and also have to be a beta tester!
6 Likes
Hello again, itās been 12 days since my original post. I wanted to check up.
I see that already 55 people have voted in that short time to show they feel the same way; thatās a number that canāt/shouldnāt be ignored.
- Can someone give me feedback on what has changed since then?
- Has the focus / planning actually been re-shifted?
- Has there been progress on the Blurry patches issue (e.g. a beta version with new āpromising directionā eric wrote about a week ago?)
Iām trying to ascertain, if itās actually going in a new directionā¦
(or just same-old, same-old with someone calming down the crowd)
Thanks in advance for any feedback on this.
@GinTonic - I completely agree with what you wrote. Going in the direction of a photo editor is a terrible idea (at least IMHO). Thatās re-inventing the wheel and there are already great products out there (both cheap and expensive ones).
2 Likes
You do realize that implementing the custom AI model feature you described would require users to have access to significantly more computing power, and very large numbers of example images.
How long does it take to develop an AI model?
Simple AI systems can be built in a matter of weeks, while more complex systems can take months or even years to build.
1 Like
What you describe could actually be provided as a (paid) service from Topaz Labs - if one wanted to go in that direction. If you are a photographer or artist who has a specific style, wouldāt it be great and even unique to have a company that lets you train an AI model specifically to your needs?
I didnāt say that this should be - the one - idea to go for. I was just contributing to userās ideas back then.
1 Like
Well, we did just jump all over them for accessing our images to create the current models without copyright permission, so there would have to be legal agreements negotiated. I expect the fees involved would not be insignificant.
What are the goals?
What is realistic, the definition of it, how does it look?
The last time a programmer told me how long it takes to train his model, he told me that it takes 14 days (24/7) of training before he has any idea if it works at all.
Thanks for the info, thatās interesting to know.
Itās always a process, gathering raw ideas, determining their technical viability (and also taking into account the profitability for the company). But I think it canāt hurt to talk about it
I didnāt mean to start a conversation about yet another old idea, it was my mistake to reference it here in the fist place. All I did back then was to give input to someones idea. But that is not what this thread is about.
My definition of realistic is that if I have an input picture (that is low res, blurry, etc.) that the output picture looks as real as it can be, without looking artificial. As if it had been taken in the first place. Keeping as much detail as can be retained, having peopleās skin and face still look like persons and not like Barbie or someone using an Instagram beautification filter to cover wrinkles (If I leave face detect with default 80 itās what I get in the current versions). At least thatās my definition of it. Yours may be different. I just know that if I process low res images in the old 1.3.2 version the resulting outputs look much more realistic than in later versions.
What interests me more is if something actually changed in the last 12 days, is there at least a tendency? (Will it make sense to renew or notā¦ Thatās what Iām trying to find outā¦)
