High res scans of 1950’s Tri-X 35mm film which are very grainy, Topaz AI2 will work well on some and actually sharpen others . . .
I pushed .jpg and .tif scans from a Nikon 4000 slide scanner of 1970s Kodachrome and Extachrome. The JPEG upscales work very well in version 2.0.0. The beta color and lighting enhancements are going to be very useful for recovering slow-ISO indoor shots.
The AI models rely on existing image data so if the original detail or resolution is poor, the app will struggle as well. I suspect the models are sharpening noise because it thinks it’s fine detail.
Try increasing the strength of the remove noise filters and see if that helps the performance.
These were 35mm tri-x that have a high grain no sharpening applied only de-noise but still sharpens!
There might be missing denoise model files. Please disable antivirus, firewall, VPN, or proxy that may be on your computer. These are known to block the installer from downloading the model files. Then rerun the installer.
If that doesn’t work, please upload one of the original scans so I can run some tests on my end and see what’s happening.
Thanks for uploading that. What seems to be happening is because the grain is much larger than what digital noise looks like, the model thinks it’s detail. Even turning on the various remove noise filters doesn’t remove it because it isn’t registered as noise in the first place.
The Upscale filter does some minor denoise and sharpening. I got the best results by turning up the Minor Deblur slider in the Remove Noise filter. Give that a try and let me know what you think.
Tried your suggestion, (thanks) but still not doing enough, Photoshop Neural “Photo Restoration” gives a good result when used as a 60% overlay on an un edited layer! Had hoped that Topaz would recognise film grain as restoring historic images from the 40’s & 50’s is an important job
I’m glad you found an alternative that works better for these photos! Our primary focus at the moment is digital photos, especially professional photographs. We do have a good portion of users who are using Photo AI for photo restoration though. We’ll continue to improve our noise removal models to treat these cases.