Hello.
I am sure many of us have seen information and had the option to sign up for the upcoming version of this software for IOS.
Will it be possible to process on the device?
Will all processing be in the cloud?
The advert mentions limitations. What will these limitations be?
Will this also be available for Android in time?
They’re doing it to sell their cloud compute. You might get very basics done on device but not much more.
Do you have inside knowledge or what’s your statement based on?
As for what you mention you really don’t need a sophisticated app, but a quite simple web-interface would be sufficient 
And, no reason to limit this to iOS, either.
I think the iOS app maybe will incorporate the „standard“ upscale /denoise/sharpen models and maybe face recovery as the ARM code is already there and runs sufficiently fast on even older plain M1 processors - so iPhone from 14 up and especially the newer iPads should be able to handle it as well.
The redefine and recover options will very likely be limited to cloud computing ( and, of course you’re right that then the iOS app is supposed to increase the revenue from that).
Modern iOS devices contain the Apple Neural Engine, capable of 35 TOPS since A17, and have similar GPU architectures to M series Macs. So, I’d imagine there’s actually quite a lot possible on device, if properly leveraged.
This is what I said. And thus the normal GP upscale models should definitely be feasible. And the code should also mostly fit since it’s basically the same ARM architecture.
BUT thinking of the absolutely abysmal performance of the new generative models (Redefine/Recover) even on better AppleSilicon desktop chips I don’t think that this would make sense to incorporate them in the mobile devices
