As a member of the beta testing group I have been running various betas of 2023 over the last month so thought it might be worth sharing some of the things I have found out and how I have come round to using 2023. I am running 2023 on an M1 iMac with 16 Gbytes and extra SSDs to spread data access, so I am on a well spec'd system. Most of my images are from M 4/3 cameras.
The upgrade has solved a couple of issues that I had been hitting in 2022
- I haven't had to do a Force Quit because PhotoRAW would not close down with "Stopping Data Services" prompt
- if automatically apply lens corrections is set in preferences and presets are saved with out lens correction being checked then preset will use the correct lens correction when applied to images shot with different lens
For 2023 itself
- cataloguing folders when Keyword AI scanning of catalogued folders does take a while, once this has been done then performance has been good
- I feel that the default 100% deblur in Tack Sharp is too high for many of my images, reducing it to 50% and adding mild Dynamic Contrast has given a reasonable starting point. (I also run NoNoise with its default set to Medium for similar reason). I have a preset with these settings and auto development settings that I find has been a good starting point.
- if both NoNoise and Tack Sharp are applied then the time to reload an image into Edit is about twice that for loading with just NoNoise. This is the price to pay for non-destructive editing where all adjustments have to be reapplied each time.
- all the new AI masking and associated workflow changes are very useful, when the masking works. Unfortunately there is a lot more training needed as some things are not correctly identified and the automatic refining of the AI mask completely messes it up. Given how sky masking has improved since the initial 2022 release I hope that things will improve over the year.
- if an AI mask is not quite right then playing with Levels in the expanded mask view can sometimes improve the mask edges.
- Sky Masking, in both Sky Swap and in the Mask Menu dropdown/right-click on mask thumbnail, seems to have retained its own AI masking and is producing much better sky masks than Super Select AI etc. Again playing with mask Levels can produce a harder edge if this is needed
- I feel the biggest masking mistake has been that the new auto AI Quick Mask has replaced the previous painted Keep/Drop version. If the latter had been retained then it would be available as another option to create a mask if the auto AI masking doesn't work, and would have meant that all the auto AI masking was in addition to what was there before. Perhaps if enough people raise this as an issue On1 may bring it back in some way, as they reinstated the Clone tool after it was initially dropped in favour of Healing Brush
- Aside from the issues with the masks themselves the other issue I hit with the new features is that there are so many ways to use them each with own quirks e.g. in Super Select AI if after creating a mask you chose the adjustment/filter by right-clicking then it gets applied immediately, but if you choose from drop down then you have to click Done before it gets added. It took me a while to understand what happened in different circumstance and learn the appropriate approach for what I was trying to do.
- some Keyword AI suggestions are a bit odd. I had to Google one that it produced on one of my images since it was a term I had never come across before. Turned out the keyword was right in that it was the name of the animal group that contains kangaroos and wallabies. Presumably hedging its bets over which it was in my image. (UK fans of ISIAHAC may have come across the Uxbridge English Dictionary definition - "Wallaby - aspiring to be a Kangaroo")
I think that most of the major bugs I hit during beta testing have been resolved. Some others remain and On1 have said that they are still working on these in advance of minor releases.
Overall I feel that the AI masking is not yet as good as suggested, and could become, but is heading in the right direction and that 2023 has improved on 2022 in many aspects.
Please sign in to leave a comment.