Current recommended SilverFast settings?

hi all.
I’ve seen different recommendations for settings for SilverFast with dust removal.

is 48-bit tiff at gamma 1.0 linear (L* Gradation) currently the recommended format?
do I need to run the Tiff Scan Prep tool with these? my tests seem to indicate the prep tool is not necessary, but the Guide states that the prep tool is required when the tiffs are saved with a gamma of 1.

I’ve tested with (a) HDR raw tiff, (b) 48-bit tiff g1.0 and (c ) 48-bit tiff g2.2
48-bit tiff g2.2 seems to produce greenish results in NLP.
HDR raw tiff and 48-bit tiff g2.2 produce the same (better) results in NLP.
I also get greenish results if I turn the scanner’s lamps up between 200 and 300 (with the intent of keeping as much of the scanned spectrum as possible within the histogram).

when exporting to HDR raw, we can choose between tif and dng. does it make a difference?

cheers
Gregory

Nikon Super Coolscan 4000ED with SilverFast 9

1 Like

hi, i’ve just started using silverfast and 48bit RAW DNG… import to LR…run the plugin>vuescan update… then whitebalance off film edge before converting w/ NLP… my resulting colours are amazing, but dust/scratches everywhere.

when scanning as TIFF (for iSRD use) i’ve set Gamma 1.0 as Nate noted somewhere… the tiff utility doesn’t really widen the histogram and the resulting scan (no white balance taken from egde) is off. when scanning TIFF at Gamma 2.2 without the utility the result is ok but nothing like the raw.

@Nate, +1 on any advice on Silverfast setting please

Hi,
I have over 80 35mm rolls to scan with a V850 Pro and while NLP looks very good without automated dust/scratch removal somewhere in the process I can’t see myself using NLP, which is a shame.
In my experiments Silverfast is very good at removing dust while Epson scan seems to be better than Vuescan.

@Nate +1 to Silverfast settings or NLP to process IR channel and remove dust.

Hi,
I scanned in some negatives with SilverFast 9 gamma 2.2 48 bit 3200 dpi Epson V850 and everything but iSRD disabled and dust / scratches disappeared (great). Hard to say how good the images are as the negatives were shot indoors without a flash. But they are better than the photo lab managed many years ago, that is they didn’t print anything! The files are 90MB so may go down to 2400 dpi.

However, I’m still on the trial and I batch converted 11 images but how do I edit in NLP after conversion?
thanks

My workflow now is scan negative to positive with Silverfast 9 AI and scan positive with iSRD for NLP both at 3200 dpi 48 bit TIFF. Seems a bit over the top but find negative to positive is a good point to set up the frames. I’m getting similar results with both but in general NLP gives me a more pleasing result.

I experimented with Sillverfast raw HDRi, DNG and tried Epson Scan and Epson Scan 2 without adding much value. In all these cases dust and scratches were an issue. With HDRi wasn’t that impressed with the Silverfast’s HDR software (its basically the same as Silverfast 9 AI but with an additional cost).

if i was super diligent i’d save the raw HDRi format too but have decided not to buy the HDRi license required as wasn’t impressed with the demo and if i’m going to fiddle with the images prefer to use better tools - NLP / lightroom / Photoshop / Pixelmator

I’m particular impressed with the colour cast removal / conversion in NLP. Basically the defaults just work unlike SilverFast where you need to fiddle about choosing films and even in some cases editing profiles. This means i can convert batches of frames and only reprocess a few in NLP.

A good job @Nate :smile:

Thanks so much @peterlappo !

And yes! I did a ton of testing on “film specific profiles” a while back (ala Vuescan, Silverfast, ColorPerfect), and had a few conclusions from that testing that heavily influenced the design of Negative Lab Pro:

1. “Evaluative” profiles (like AutoNeutral and AutoWarm) typically outperformed “fixed” film profiles.
2. Whether evaluative or fixed, it all needs to be integrated into a single tool to avoid a “stacking” effect. For instance, some tools enable you to use both an autocolor algorithm and a fixed film profile at the same time, which stacks the effects and leads to bad color.
3. Whether evaluative or fixed, it should be easy to see and adjust exactly what it is doing. This is one thing that drove me nuts about Photoshop plugins I tried in the past - pretty much every photoshop plugin uses one or more “autocolor” adjustments in an early part of the conversion, but these adjustments end up “fully baked” in the output - there is no transparency to color adjustment is being applied, and no ability to change it in the early, crucial state in the editing pipeline.

I’m trying to find a better way to do this… the Tiff Scan Prep utility is not ideal and there are just too many caveats to it. I’m working on a method that could avoid this, and that could make for a better workflow for those wanting the IR dust removal embedded into their Silverfast scans.

The other huge benefit of the Lightroom / NLP method is non-destructive editing. I quite like Apple Photos for that reason.

Agreed. It gets very confusing to the user when there are too many ways of editing most of which are not very well described so you start by experimenting but eventually give up especially when the effect is marginal.

I had a play with the pixelmator ML (machine learning) enhancement tool on some early results from Silverfast with good effect and it removed a mauve colour cast. I was using it as Photos plugin but haven’t tried with Lightroom.

I’ve set SF9 to gamma 2.2 with Linear gamma. I found it to be the best setting. gamma 1 requires pre-processing before running it through NLP, and other image editing apps can’t handle the images either. g2.2 is much easier to work with.

my Nikon Coolscan 4000ED scanner allows you to adjust colour channels during scanning. I try to remove as much cast as possible from slides during the scanning process. the cast is generally an artefact of ageing. trying to apply big colour changes later can be very troublesome.