Can you add natural photo grain/noise control or match existing photo grain/noise ?
For example after face recovery the face looks too smooth and pop out compared to the rest of the photo - it looks unnatural . I want to mask the face and to add some natural noise/grain so the face match the rest of the photo so the entire photo become looks more equal and natural .
Maybe it could be something similar to this SD Film Grain Noise but even better if it could work with mask too .
Let’s be clear, though; this so-called “filmic” “natural-look” ect. about adding grain, ect. is ONLY an artistic choice… period, and to hide compression artifacts, ect. that lazy creators failed to professional edit correctly.
When film grain were used back-in-the-day… it was due to the creators very limited optical choices / technology and not because most creators wanted that “look” into their final delivery.
There’s nothing “natural” by witnessing human teeth that has “running ants” between them, clouds/sky looks dirty with sprayed on fine pepper, white/bright objects looks like Dalmatian dogs, ect.
Grain is just an illusion (a very poor one, IMO) that we’ve all witness for decades and it’s just weird to some folks to see material without there sprayed-on pepper / running ants in as well.
I get what you mean for the restore faces part of photo AI. I’ve been using it myself for restoring some glass slide scans from around the1920’s and camera tech as you can imagine was basic with a lot of grain. Photo AI is amazing but the restore faces is SO GOOD it looks like “Uncanny valley stuck on” faces at times. You can dial back the effect but that kinduv defeats the object. I ideal is for it to “sit” in the same conditions (grain) as the rest of the image.
Picking up on what you said about SD Film Grain Noise there was a greta plugin for Final Cut Pro a few years back called Neat Video that “learns” what was noise in a VHS video and removes it and preserves details. The User selects a “patch” of the video image that is neutral; without detail or contrast so the plug in can sample the grain/noise apply the filter to remove it.
What you want is the opposite; where the algorithm understands the texture, size and strength of the grain and applies it to the recovered faces from a sample elsewhere in the video that’s just grain. The advantage of this approach is it’s the ACTUAL grain of the photo, not an arbitary made up approximation. All grains are not equal! A couple of sliders to reduce the effect in lighter areas and shadow… Bob’s your uncle.
I pretty much do that with the Camera Raw filter in Photoshop, adding grain to the face areas (after a feathered selection using Color Range to only select the midtones; tail off the effect in extreme highs and shadows) and messing with the sliders to “knockback” the clarity so it fits. It’s pretty successful.
(Hope that helps you in the meantime!)
Honestly, in terms of photographic restoration, this is something TopazLabs should look into as it will make the software even better for historical recovery. It’s all about tricking the eye!
