Most important problem from the beginning (at least about 2 years): I still can’t define/redefine frame for “recover faces” and automatic framing leaves ears, neck, part/piece of hair NOT RECOVERED !!!
There is still problem with recovering scanned photos from structured paper (canvas, luster).
If there is technical limit for 32000 px - I understand, but I still don’t understand limit of 6x upscale.
The current AI models are not trained on scanned images. New models would probably need to be created for that purpose. If that is something you particularly want then you can create a request in the ‘Ideas’ section of the forum for users to vote on.
That is a suggestion you can already vote for in the ideas section:
The AI models actually only upscale to 4x. So when you select 6x upscale, the AI model does a 4x upscale on your image and then a standard upscaler (Catmull-Rom I think) takes it the rest of the way.