As we gather feedback and test different images with the Recover model, I came across a few nice examples where Recover v3 performs very well compared to Recover v2. In general, I would say it works better with fur and feathers and renders a more natural look. From left to right:
Original file for those who would like to test.
I’ll share more later 
A couple more iPhone photos shot yesterday for my Extreme Enhancement project while out and about. Both used Redefine>Creative @ 6X.
Having some weirdness today with 1.1.2 (Mac M1 Ultra, Sequoia).
After updating the app and doing a successful group of Redefines (below), new cloud uploads/renders are failing (this is after a crash).
For this second image I restarted GP after the crash and clicked on the Recent photo icon and got photo 1 instead, which was not listed.
Original iPhone windshield photo from this morning, reduced:
6X with Redefine>Creative
Restarted the app again, I think we are back in business:
Quick windshield shot:
4X Redefine>Creative:
I like the third photo (cauldron) redefine creative-6x. Such very picturesque mountains with snow and clouds below the peak.
Thanks, in winter we get some drama going on up there, otherwise it’s just dry dirt and bland sky 
Hi, when will there be a new update or beta? There hasn’t been anything since the FEBRUARY version.
If you reboot the computer and try again, is it working? Cloud rendering failing could be because it’s down as well. I know AWS had issues yesterday.
We don’t have a beta planned for now.
In terms of ‘best’ differences between cloud and local models, it’s the Recover v2 model, I’d say. In the following GIF animation, first is the cloud-processed image, then it’s local-processed, on my 3090/Win10/Gigapixel 1.1.2, right this moment. I switch between them with a mouse roll in the XnViewMP viewer:
Local version seems to output a tad better Akari’s face.
Imagine giving this kind of output from this. Yes, that’s right.
Recover v3 cloud low isn’t doing great here though - too much of the JPEG mess in the source image. The mess increases as I went to medium and high.
An original for any brave soul to try.
There is still a small grain visible on the one on the left. It is less visible on the source image.
We released v1.1.3 today, please write in the new thread if you have issues.
