a lot of people in this forum have written, that they have problems with the new Sharpen V2 or that V1 was simply better in many situations.
Just one example:
In previous Photo AI versions V2 and V1 coexisted along side - users were able to choose, which model they wanted to use. In current versions V1 was simply deleted and now users have no more choice.
Many of the users do not understand this (political) decision, since clearly it is not a technical decision (as it coexisted without problems in the past; the V1 model was well trained and proven for many releases - so there should be no downside in letting it remain in the application). There’s people that are now forced to stay with an older version of Photo AI, since that is the only way they can continue to use that model:
Therefore, as many before me have already asked and also I myself have asked repeatedly in other threads: Can you please bring back Sharpen V1 in the next releases (and keep it in future versions)? You can keep V2 as default or even better - let the user decide in the autopilot preferences, which model is preferred. But please also return V1.
What you see here is 100% not sharpen V2 alone, it can’t do that at all, at least Denoise strong is still active.
If you want to vote for something, vote for the autopilot to be switched off as default, It should just ask people if they want the autopilot on or not at first startup, I think this solves 99% of the problems people have with PhotoAI., it’s what makes PhotoAI look bad.
I did vote for that Didn’t they already implement that? (I think only the annoying popup is still there suggesting that it is still auto-piloting).
I wasn’t suggesting that it’s all Sharpen’s fault, the various threads on the matter pertain to different problems people have with Sharpen V2. All I know is that sometimes Sharpen V1 (in conjunction with other filters) seems to offer better results for some users, in some cases.
Therefore I see no reason NOT to bring it back into Photo AI. I mean it is already there (on their file servers for the installer download), it was there in parallel V1 & V2 until recently (when it got replaced). Why rob users of something that works for them and has no technical reason that warrens removal?
My guess about their reasoning would be that cutting out the V1 option makes the code smaller, faster, easier, and cheaper to maintain long term. At some point in the future, when the subscription volume and new sales drops off, they will want to surplus most of their current staff, and legacy stuff like this (and DeNoise, Sharpen, and Gigapixel AI) will be a burden on their remaining resources.
Being a software engineer for over 2 decades myself, I can only say it all depends on your architecture…
Of course the code can be laced with lots of if/else conditions and one can build up mega-complexity in such a way, but I think Topaz Labs knows better than to do that…
You can do so much with inheritance and the use of design patterns (like composite, etc.).
The way I picture it, and I would hope they built it in such a way, is that the models are just one fragment. And that they can add/remove them easily (load / unload). They might differ in signatures (the newer models having additional parameters that previous ones didn’t have), but once you load such a module it is good to go.
And the fact that V1 and V2 already coexisted says that it works this way.
Additionally the models are proven. And they are (still) located on their Internet file server. They have to maintain that anyhow (as part of support of older versions, as we all can still install older versions).
And lastly, since there seems to be little to no integration/function testing thus far, it won’t matter if an additional model is in there or not; if one were to program tests one would focus on the current implementation (V2) of course.
I might also be totally mistaken and the code is chaos and mega complex. But I wouldn’t think so. It’s not like they just started programming
Just my 2 Cents
Personally: what bugs me the most is getting no response whatsoever on the request to bring back V1. I’m not talking about this new thread in ideas. There have been many requests to bring back V1. It would be OK if they said, yes but, or not but… - but not responding at all to users requesting V1 back and having trouble with V2 is just not acceptable.
Yes it would be, if they were to actually implement it. I already voted for your idea and left a comment a while back.
But I’ve had ideas which now have reached 97 votes and not much has changed, so keep your fingers crossed…