Great. I hope new speed of Denoise and Sharpen will also come to Gigapixel.
Why when i upscaling one image, GPAI remember settings but when i upscaling 2 or more image then GPAI do not remember file format and color profile ?
It will - weāve got a few more things to iron out but itāll be pulled in for our big April release of Gigapixel (slated for the 21st). If youād like to check it out before then, we always need more beta testers
What do you mean?
Great. How i can be beta tester?
Added you to the beta group. The Gigapixel AI beta with the updated AI engine + new AI models will be going out on the 7th of next week, so keep an eye out!
Great. Thanks
That sounds good, Iām really looking forward to it.
Can I get into the beta as well?
You should be in both the Video & Photo beta groups now Let me know if you canāt see the āPhoto Beta Testingā section of the forums.
One image: GPAI remember last settings:
Multiple Image: GPAI not remember last settings:
Thanks. BTW, is the next version going to have similar results between GPU and CPU? I tested out the latest beta and CPU is still giving me an overall sharper image.
So I think I was in violent agreement with youā¦ that it should be a Pref or other Setting one could establish. Based on what I said above. I believe itās MAI that gives users the option to write to a new layer or over-write the launch layer. That same type of capability/choice would be good in all the plugin products. Iām always concerned Iām going to torch something I shouldnāt if I allow overwrites and it makes it horrendous to back up (vs delete a layer) to step backward through steps if the changes get baked in. If someone knows for sure theyāre ready to jump off the bridge and commit to a change it likely would be preferable to have fewer layers and potentially less file memory genād.
Agree with all of this. The reason Iād like this option is because Gigapixel is usually my very first action after opening an image so itās not going to interfere with work history. If Iām working on a lot of images, having to manually delete the original layer every single time is annoying.
That is never going to happen. I suspect is its a hardware limitation.
Using trial version of v5.4.5 on Win 10. No issues with crashes, etc. but I am very disappointed in functionality for my application. I was hoping for truly āintelligentā upscaling that would recognize pixel jaggies on a low-res image and eliminate them on upscaling. No such luck. I just see sharper jaggies. Useless to me in all the Gigapixel AI modes.
A true AI might recognize, for example, the smooth curve of a fabric drape and eliminate the pixel problem. Not happening here.
Iām upscaling some Cyberpunk 2077 fanart I found a while back and the results that it provides are kinda disappointing.
Iām not sure why the AI is applying so much red to the faces after trying two different models. Itās not how the original looks at all. Am I missing something or is this normal?
Youāre upscaling a jpeg that was compressed with chroma subsampling (meaning it had loss of color information, so I believe itās trying to undo that). Also, in my experience with the software, anything above 2x resolution on digital artwork will have poor results, thatās just a limitation of the AI. 2x should be more than enough for that image anyways.
Retried following the 2x option. Much better. Iāll keep that in mind for anything else I do.
7th of next week? Did you mean 7th of next month, i.e. April 7th? Anyway I am excited to hear that the faster AI engine and retrained models are in testing, woooohoooooooooooo!