RoboScan: Lego+Raspberry Pi-powered analog film roll scanner
github.com
github.com
For anyone wanting to do this, you will need to invert your b&w or colour negatives with software capable of processing RAW files (lightroom, darktable, rawtherapee) white balance the negative, invert and then colour/exposure correct the final image. I use darktable for this and the latest version has a real nice plugin called negadoctor that can give you a great starting point. For lightroom, NegativeLabPro is the goto tool and works in a similar way to negadoctor but is better at producing the initial image.
The outcome quality of both processes is identical and great quality, but the DSLR repro way is a little bit more cumbersome. It takes me longer to get a really good result with a lightbox than with the dedicated film scanner. Not by long, but it adds up if I have a couple of 120 film to repro.
I shoot repro with a Nikon D700 in RAW using a 105mm macro lens on a regular stand, with a bubble calibrator. The lightbox is a small cardboard box spray painted in black, with flippable/removable cutouts to fit 6x4,5 up to 6x7. I use an old discarded iPad mini, setup with a white image on max brightness as backlight.
Using Capture One (my preference of editing suite), I connect my camera with a USB cable and import directly into the editing suite. No need for importing of a CF/SD card. Smooth process.
Time wise, the speed of snapping a photo is also mostly outweighted by the time needed to post process the image (neg invert, white balance, colour correct) versus the slow scanning speed of a flatbed negative scanner. Never used the cheaper 35mm scanners you can buy, but from what I’ve seen they're just the dslr approach without the quality.
I think it comes down to two questions. 1: Do you own a dslr with a macro lens? 2: Do you know how to use raw editors like lightroom or darktable?
If yes to both, then dslr scanning is easily the cheaper of the two and is most likely the best route to go to test the process.
to save time, i have saved some presets for colour correcting using a single curves layer. since i end up tweaking every scan anyway, it speeds down the workflow by cutting down on the process.
Also if you have that many negatives, every step you need to take is one step too many. My frames came from the camera without any borders, so no cropping needed. Don't put the negatives in the "wrong" way. Base vs. emulsion debate is just that - a debate with miniscule differences.
I had a foot-switch for release so that you can use both hands for the camera and focus. Then you'll achieve ~5 minutes for a full 36 frame roll of (strips of 4-6 frames) film.
Self plug: I was annoyed by the lack of indicators: when it's a picture or a youtube link that I created https://mhasbini.com/blog/embed-youtube-in-github.html
E.g. of before & after: https://github.com/bezineb5/RoboScan/pull/2
You can get 90% as good results by putting the negative on a slide viewer (more commonly sold for tracing these days) and taking a picture with your digital camera. Will you extract every bit of information that the film captured? No. Will an amazing picture look amazing when digitized in this way? Absolutely. So it's worth considering if you already have a digital camera and are just getting started!
I have taken pictures of negatives hanging to dry with my phone, inverted them in Photoshop, and posted them to social media with good results. You don't need expensive equipment and to spend hours getting things perfect if you don't want to.
Reminds me of the mongoose by cameradactyl https://www.cameradactyl.com/mongoose
Perhaps direct projection onto the sensor (rather than using a lense) could make sense as well.
Scanning movie films used to be ridiculously expensive (and it shouldn't).
It was expensive because it was pushing the boundaries of what was possible.
the northlight (https://www.filmlight.ltd.uk/products/northlight/overview_nl...) which I had the pleasure of working with the prototype of is a spectacular machine.
Firstly it has really good registration, which means that in theory you should be able to scan the same film twice and it'll look exactly the same. Not only that its gentle, so you can scan really fragile stuff.
Add to that it over scans the living snot out of the film, plus an IR pass to get rid of the dust.
Also its not a square CCD, its a linear CCD that is scanned across the plate.
Has anyone seen a utility to recognize and deskew prints on a flatbed scanner? Found some interactive apps, but nothing that is automated to work in a pipeline like shown here.
There are 16 bit TIFFs and JPEGs, along with a zoomed in crop (PNGs bc i wanted to have them be the same size for a side by side comparison). For the 2 scanners, these were scanned fully "raw", without any software colour correction or sharpening (FlexColor for the Imacon, Epson Scan 2 for the V600). The camera scan is using my 2012 Fuji X-E1 with a Micro-Nikkor 55mm @ f/8, and the V600 negative carrier, with an iPad as a light source (there's some barrel distortion; i'm still ironing out my rig). Final results corrected by my meatbag eye in Affinity (I had to increase saturation of the V600 scans to match the other two), hence the slightly different colour rendition, but the difference in grain resolution and overall detail should be pretty clear.
From a quick glance over the code, it seems to scan over the film roll and takes photos using the camera.
But I haven't really found much logic for post processing. You would need to adapt the colors in some way, right? It's a negative, so inverse them. But then they still need to be adapted in some way. Actually this was for me the most unclear aspect about such a project -- how to deal with the colors.
Darktable also supports Lua scripting, maybe that could be used to automate a workflow.