No recognition of half-tones

I find high-compression is the best, but only by a bit. More than 2x is useless.
halftone for topaz

Hi John,

Have you been able to test other images for better results?

Our Ai models are trained using digital images from digital cameras so scanned images or old photos from negatives tend to not be as consistent result-wise.

Screen grabs are often the only way of getting images off the web; many old photos can only be found in newspaper archives. Both are rectangular grid samples. An auxiliary processor for them similar to face recognition would help a lot for researchers like me.

This is the top image scaled up by 2 x via the new Recovery Beta:

And this one was smudged via gaussian blur level 1,0 in Photoshop before running the new Recovery Beta:

Thank you, Imo. As Topaz says, it’s designed for basically good digital camera images :cry: