Conversely if you have detected a subject you might like to just change the lighting of everything else.
Plus I think there’s an option (for the future) to be a bit-iPhone like and detect assorted features (Sky, Subject, Skin, Hair, etc.) and apply separate processing to each (for all available processing that makes sense, so not resizing ).
Here’s an example of using iPhone ProRaw semantic masks FYI (which the phone detects and includes in the ProRaw file) and is the sort of thing an AI tool should be good at vs. a conventional one:
Hi
Thanks for your continued effort to improve Photo AI. Unfortunately, the upgrade to v2 for me is somewhat disappointing. I do not need tools for lighting or color balancing. People, who work with RAWs and TIFFs from C1 or Photoshop do not need these tools – they can do it in a more subtle way in these applications. The speciality of Photo AI is improving images, that lack information (resolution, sharpness) or have too much noise. There you have done a great job – and you could stop with 1.5.5 with a slight but. I have always asked for an option to add grainy structure since it is the only way for getting better looking images from DJI drone DNGs. Unfortunately this request has never been heard. You say that recovery of the original information should do the job, but it just can’t if there is no such information (relatively low res image from the drone - 48 MP but on a very small sensor). This is reason why I still have to use Nik Color Efex to add grain. I would have been happy if v2.0 would have finally brought this add on, which has a close relation to what the speciality of Photo AI is – but the new beta tools definitely do not
Yes, it’s still happening… only for RAW files. I’ve switched to jpg/tiff. I have a video review of 2.0 that shows the issues and tips on my editing workflow.
The app has been continuously crashing when I combine upscale + sharpen/denoise + face recovery
Tried on a few pics, crashes after each face recovery turn on
extremely slow export times for PNG files, the 1.5 version was much quicker
Edit: Another thing - I just cancelled a process to change a few settings, but seems like it exported both versions anyway - with the first settings and the new settings.
Photo AI v2.0.0
Unable to load an image without the program crashing.
win 11
13th Gen Intel(R) Core™ i9-13900K
NVIDIA GeForce RTX 4080
Suggestions greatly appreciated.
A pretty dramatic recovery of shadow detail from a JPEG shot several years ago (2012 Nikon Coolpix non-DSLR). I did not need to crank the slider to the max. You might call this “Southern California on a good day”
Other than this new capability (which is most welcome!), I would ask for a Levels panel or even Black Point/White Point sliders to be added to the existing Adjusting Lighting panel for the next degree of control – because sometimes you need to adjust each independently (as in PS Shadows/Highlights).
A pleasing salvage of a 20+ year-old Kodak 1-megapixel image! I find Photo AI absolutely necessary for making my decades of photos (film, scan and early digital-sourced) usable in the modern world.
I’m new to this but wanted to point out a bug when using Topaz v2 after editing
a Sony A1 photo in Adobe Camera Raw using the new denoise feature in that app. Weird artifacts appear when using sharpen and denoise together. To wit:
Balance Color is not useable as you can see here, in fact this is quite an extreme example because of the underlying colors in the image and the result is so far OFF … the original image clearly shows in the navigator …
PAI V2, at least for me, is a an improvement. Raw noise models have better utilization of the GPU. Raw V2 models are much better. If used as a plugin, I really don’t see any use for the “adjust lighting” or “balanced color” controls. Same adjustments can be done in PS or LRc or …
).
