Show your Bird Portraits

[P-visie]

No argument re the best being a good image to start with. I find with Topaz that I have to experiment with the different modes and the slider level to get the best result with the least artifacts. Note that they do update the models almost weekly (so you do need a wideband connection to keep up) and some of the models are getting pretty good. AI will (almost by definition) add detail (i.e. artifacts) that were not in the original image exactly as presented. That is also the virtual case when a human eye looks at a noisy picture and interpolates a better result. For most work, I think the AI programs will be able to produce much improved images from noisy originals that meet the test of a critical eye, but they are clearly going to be a challenge for legal images in court cases (where exacting detail is critical). I also dislike the subscription model, but It seems that all software is eventually going to go that way. Kind of like phone companies charging a flat fee every month instead of billing by the call. They want to know what their revenue is going to be. The model I really dislike is the one where the processing (and backup) is in the "cloud". Topaz at least gives you the choice of not processing in the cloud (i.e. on their computers). Apple and Microsoft are really bad about that.

I’ve given up on Topaz: it took them 3+ months to support the R5Mk II, their support persons lied to me why it took them so long, they need to improve their software quality and quality assurance (the reason for their frequent updates) and their pricing is crazy: at the introduction of the subscription model, their basic subscription price was 50% higher than the Adobe photography plan.

 

[AlanF]

No argument re the best being a good image to start with. I find with Topaz that I have to experiment with the different modes and the slider level to get the best result with the least artifacts. Note that they do update the models almost weekly (so you do need a wideband connection to keep up) and some of the models are getting pretty good. AI will (almost by definition) add detail (i.e. artifacts) that were not in the original image exactly as presented. That is also the virtual case when a human eye looks at a noisy picture and interpolates a better result. For most work, I think the AI programs will be able to produce much improved images from noisy originals that meet the test of a critical eye, but they are clearly going to be a challenge for legal images in court cases (where exacting detail is critical). I also dislike the subscription model, but It seems that all software is eventually going to go that way. Kind of like phone companies charging a flat fee every month instead of billing by the call. They want to know what their revenue is going to be. The model I really dislike is the one where the processing (and backup) is in the "cloud". Topaz at least gives you the choice of not processing in the cloud (i.e. on their computers). Apple and Microsoft are really bad about that.

My experience is that if the soft image does not have much detail, then the AI models do a reasonable and sometimes excellent job improving acutance. But if there is blurred detail, they hallucinate to deconvolute that detail.

 

[AlanF]

A couple of shots from today using the 2xTC on the RF 100-500mm and R5ii at 1000mm f/14. A Corn Bunting, that was hiding at the top of a tree in strong winds that for a couple of shots parted the leaves so I could see it, and a Grey Wagtail on lotus leaves. The good thing for me this time of year with this combo is I can use the RF 100-500mm for shots of dragonflies, and put the TC on both for that and distant small birds. For comparison, which is of interest to @Maximilian as he is testing his lenses, there are shots of a Grey Wagtail in the distance (18m) at 500mm and 1000mm from today with the third being from the 200-800mm a week ago, a little closer.