I am trying to lockdown my workflow for digitizing, using Topaz Photo AI.
I currently shoot with a Nikon Z6, Nikkor 55 2.8 AIS macro, on a Negative Supply rig.
I’m liking the denoise and sharpening I get when processing a RAW file, but they asked me an interesting question about why shoot in RAW in the first place. I’ve pasted their comment below. Thoughts? Why do we shoot in RAW, presuming white balance and dynamic range is sufficient in JPG? I’ve always thought of it as standard practice, but haven’t really thought about “why.” I also have never heard of RAW noise as compared to JPG noise.
“What I need help understanding is the reason why you would Scan/shoot your files in RAW, and this adds RAW NOISE, that is not present in your original scan, only to add a extra step to have to remove it? That is not needed. I would strongly recommend to shoot in JPEG, and avoid adding RAW Noise. The JPEG produced is hi-res and high in MP count. This would be different when you shoot actual pictures, yes, shoot in RAW, but to Scan with your camera, there is no need to shoot in RAW.”
As far as I’m aware nothing about shooting in RAW adds any noise, but most cameras automatically apply denoise and sharpening when shooting JPG. I mean technically the camera is always shooting in RAW, it’s just converting the file to a JPG after.
The compression from saving in JPG will destroy data that will be unrecoverable. For the sake of archiving digital copies of my negatives, that is not something I desire.
If you are inverting with NLP then I imagine converting a JPG will affect it since NLP was designed for working with the linear data from a RAW file. If you look at the histogram before and after NLP’s inversion you will notice that the movement of data is quite significant.
On another note, if your scans are coming out so soft and noisy that you require significant measures such as AI tools to fix then I think something is very wrong with your scanning setup or technique.
The scans are sharp enough. Photo AI makes them sharper in a way that, in my opinion, does not look artificial or anything.
Thanks for the response. I’ll likely drop Photo AI from the first part of my workflow, and just have it do sharpening at the end. I still like it for that.
This is my “rule of thumb” based on my experience. JPG is a finished image, do not try to edit it. RAW is a digital negative that contains all the dynamic range and detail the camera is capable of. Sometimes a JPG is all that is available, it will have limited color correction and some software cannot do noise reduction without softening/bluring the image. Photoshop, Lightroom, and DXO PhotoLab require a RAW file for the best noise reduction. Topaz has great software, I use 2 of their products Video AI and Gigapixel.
PS Since you are digitizing film negatives, you can always reshoot if a JPG can’t be fixed. It’s only more time!
Coming to this late but I tried using Topaz Photo AI (for upscaling) and I believe it is just not trained on how to understand film grain. Sometimes it does alright but usually it came out terribly. So I stopped trying to figure it out.
This is anecdotal of course. I would love for someone to show me how I am wrong but to achieve larger print sizes I ended up with a pixel-shift camera (you can see my post history for that).
I found the built-in upscale in Adobe Camera raw did a better job - but still not up to the standard I wanted, so I went the hardware route.
I am not getting good results from 35mm at all. MF is not bad IMO, and LF seems to benefit quite a bit. So as the grain lessens, Topaz seems to be beneficial, in my experience. I would love to upgrade my camera but it’s not in the cards for right now.