WEBVTT Kind: captions Language: en 00:00:00.070 --> 00:00:04.780 I learned how to use a camera with these, a pair of Canon cameras given to me by a close 00:00:04.780 --> 00:00:06.170 family friend. 00:00:06.170 --> 00:00:10.299 This one in particular, the Canon F-1, is delightful to use. 00:00:10.299 --> 00:00:14.750 Of course people familiar with the Canon T90 will know that it is arguably the best manual 00:00:14.750 --> 00:00:19.550 focus camera Canon ever made, but the vintage charm, clunky feel, and primitive operation 00:00:19.550 --> 00:00:23.160 of the F-1 made it my preferred go-to because I’m weird and eccentric. 00:00:23.160 --> 00:00:28.650 I first got my hands on these in 2007, when a digital SLR was well out of reach for me. 00:00:28.650 --> 00:00:32.880 At the time, ordinary print film could still be purchased pretty much everywhere, and Costco 00:00:32.880 --> 00:00:35.460 would develop a roll for $1.89. 00:00:35.460 --> 00:00:40.320 Much cheaper than spending a thousand dollars on a decent DSLR and lens to go with it. 00:00:40.320 --> 00:00:44.480 Anyway, the one thing I needed was a competent film scanner, as I planned to just pay to 00:00:44.480 --> 00:00:48.760 have the film developed, then scan it at a high resolution--a roundabout way of having 00:00:48.760 --> 00:00:49.980 a digital camera. 00:00:49.980 --> 00:00:54.460 I ended up buying a refurbished Epson scanner, and it had a feature I didn’t really understand 00:00:54.460 --> 00:00:58.710 at the time, but which was really cool: Digital ICE. 00:00:58.710 --> 00:01:02.290 Digital ICE may sound like the most nineties things ever, but in fact the ICE standards 00:01:02.290 --> 00:01:04.420 for Image Correction and Enhancement. 00:01:04.420 --> 00:01:09.090 The system used in film scanners was developed by Kodak at their Austin Research Center, 00:01:09.090 --> 00:01:13.400 though at the time it was referred to as Applied Science Fiction which is a much better name. 00:01:13.400 --> 00:01:18.320 I’m not sure exactly when the system was released but it probably was in the nineties, 00:01:18.320 --> 00:01:24.240 as this PC magazine article from 1998 seems totally amazed at the new Nikon Super Coolscan 00:01:24.240 --> 00:01:25.240 2000. 00:01:25.240 --> 00:01:28.480 Let’s join together in a brief moment of silence for Kodak. 00:01:28.480 --> 00:01:31.520 You had it all, and then you just… 00:01:31.580 --> 00:01:32.840 Yeah. 00:01:32.840 --> 00:01:34.659 So, what does Digital ICE actually do? 00:01:34.659 --> 00:01:39.300 Well, it’s a dust and scratch removal feature that works almost miraculously well. 00:01:39.300 --> 00:01:42.520 It can make an image go from this, to this. 00:01:42.520 --> 00:01:46.450 Some absolute purists will insist on never using it because it softens the image slightly 00:01:46.450 --> 00:01:51.960 and ...ehhh... but most people with this attitude just scare people away from photography altogether. 00:01:51.960 --> 00:01:52.960 Side-note 00:01:52.960 --> 00:01:57.270 Photography enthusiasts tend to get hung up on specs like color accuracy, sharpness, compression, 00:01:57.270 --> 00:02:00.020 file format, and more to a pedantic level. 00:02:00.020 --> 00:02:04.490 I understand that professional photographers want and need the best equipment and will 00:02:04.490 --> 00:02:09.310 work to find the best methods, but for amateurs like myself this sometimes works to create 00:02:09.310 --> 00:02:11.140 lingering self-doubt. 00:02:11.140 --> 00:02:15.010 Forum threads arguing about which camera is best, which settings you should use, why you 00:02:15.010 --> 00:02:20.180 should NEVER use jpegs, etc. can end up making those reading the thread worry that they’re 00:02:20.180 --> 00:02:24.870 not choosing the optimal methods, and that somehow their work is worse for it. 00:02:24.870 --> 00:02:25.910 My advice? 00:02:25.910 --> 00:02:27.260 Do what you think is best. 00:02:27.260 --> 00:02:30.160 Did the automatic color correction make it look better to you? 00:02:30.160 --> 00:02:31.720 If yes, go ahead and use it. 00:02:31.720 --> 00:02:36.000 Don’t go down the rabbit hole of chasing technical perfection at the expense of exploring 00:02:36.000 --> 00:02:37.000 your own creativity. 00:02:37.000 --> 00:02:39.420 If it looks good to you, that’s all that matters. 00:02:39.420 --> 00:02:43.050 With that out of the way, let’s get back to the subject at hand because it’s freaking 00:02:43.050 --> 00:02:44.050 cool! 00:02:44.050 --> 00:02:45.730 To show you, I’ve scanned some film. 00:02:45.730 --> 00:02:49.200 This long-lost set of negatives has been sitting around, gathering dust. 00:02:49.200 --> 00:02:52.430 So, when scanned, they looked like this. 00:02:52.430 --> 00:02:56.180 Dust on the film will appear white because the dust blocks light, which the scanner interprets 00:02:56.180 --> 00:03:01.420 correctly as dark spots, but as the film contains a negative image, it has to be inverted, and 00:03:01.420 --> 00:03:04.880 these dust spots get inverted to bright white when rendering the image as a positive. 00:03:04.880 --> 00:03:09.960 But, when you scan them with Digital ICE, they look like this. 00:03:09.960 --> 00:03:10.960 Pretty crazy, right! 00:03:10.960 --> 00:03:13.510 But what’s even cooler, is how this works. 00:03:13.510 --> 00:03:15.070 First a bit about the scanner. 00:03:15.070 --> 00:03:16.960 This is not the same scanner that I started with. 00:03:16.960 --> 00:03:20.980 In fact, this was one of those rare Goodwill finds where you ask yourself if it was donated 00:03:20.980 --> 00:03:21.980 by mistake. 00:03:21.980 --> 00:03:27.090 It’s a $210 dollar purchase today on Amazon, and it cost well more than that at the time 00:03:27.090 --> 00:03:29.140 I picked it up for this paltry sum. 00:03:29.140 --> 00:03:33.780 And it’s a substantial upgrade from my old one as it uses LEDs as its illumination source, 00:03:33.780 --> 00:03:36.200 and not cold cathode fluorescent lights which require warm-up. 00:03:36.200 --> 00:03:38.820 You’ll see why that’s so important later on. 00:03:38.820 --> 00:03:42.930 This is a flatbed scanner like any ordinary document scanner, and when scanning a photograph 00:03:42.930 --> 00:03:47.830 or a piece of paper, the scanner uses two rows of LEDs to illuminate whatever it’s 00:03:47.830 --> 00:03:52.430 scanning, and between them lies a single row of really tiny light sensors that build an 00:03:52.430 --> 00:03:55.680 image as the scanning head dutifully travels below the glass. 00:03:55.680 --> 00:03:59.870 But this doesn’t work to scan film, because light needs to travel through the film and 00:03:59.870 --> 00:04:01.340 not reflect off of it. 00:04:01.340 --> 00:04:05.370 What separates this scanner from an ordinary one is a second light source located in the 00:04:05.370 --> 00:04:06.810 lid of the scanner. 00:04:06.810 --> 00:04:10.590 This functions as a travelling backlight, and it follows the scanning head to allow 00:04:10.590 --> 00:04:12.410 for scanning film. 00:04:12.410 --> 00:04:16.181 Film scanners are also separated from their document brothers by a very high resolution 00:04:16.181 --> 00:04:17.650 scanning array. 00:04:17.650 --> 00:04:22.970 This scanner’s sensor is capable of resolving 6,400 pixels per inch, meaning it can produce 00:04:22.970 --> 00:04:28.440 a theoretical image with dimensions over 6,000 by 9,000 pixels from a standard frame of film, 00:04:28.440 --> 00:04:31.090 or an over 54 megapixel image. 00:04:31.090 --> 00:04:35.700 In nearly all cases this is overkill to the extreme, and it’s likely that this consumer-grade 00:04:35.700 --> 00:04:38.890 scanner’s stepper motors can’t reliably move in small enough increments to actually 00:04:38.890 --> 00:04:42.680 resolve that resolution in the Y dimension anyway, but still. 00:04:42.680 --> 00:04:43.680 Impressive. 00:04:43.680 --> 00:04:48.370 For the record, I usually scan at 3,200 DPI which is able to resolve the grain of many 00:04:48.370 --> 00:04:49.370 films. 00:04:49.370 --> 00:04:50.720 Now, to the fun part. 00:04:50.720 --> 00:04:55.240 When you use Digital ICE, the scanner actually scans each negative or slide twice. 00:04:55.240 --> 00:05:00.260 You’ll hear it get into position, make one pass of a scan, then the backlight and scanning 00:05:00.260 --> 00:05:01.880 head return to their parked position. 00:05:01.880 --> 00:05:05.350 But right when they stop, you hear a click. 00:05:05.350 --> 00:05:10.760 That click is the light source switching from your standard issue white light, to infrared. 00:05:10.760 --> 00:05:15.180 You can see this with a smartphone camera--many smartphone cameras detect infrared light and 00:05:15.180 --> 00:05:17.040 render it as a purplish glow. 00:05:17.040 --> 00:05:19.800 This is also handy for seeing if remote control batteries are any good. 00:05:19.800 --> 00:05:24.860 You see, the film itself is mostly transparent to infrared light, but dust on the film will 00:05:24.860 --> 00:05:27.070 block it from coming through. 00:05:27.070 --> 00:05:30.380 When the scanner makes it way through the second time, it will see an image that is 00:05:30.380 --> 00:05:34.990 completely empty except for where dust lies on the film. 00:05:34.990 --> 00:05:39.230 Scratches and oils from fingerprints will distort the infrared light, and the scanner 00:05:39.230 --> 00:05:41.520 can detect this distortion, too. 00:05:41.520 --> 00:05:45.210 The result is that on the second pass, the scanner will produce an image composed only 00:05:45.210 --> 00:05:49.640 of where scratches, dust, or general defects are located on the film. 00:05:49.640 --> 00:05:53.840 After it’s scanned the second time, software inside the scanner creates a composite of 00:05:53.840 --> 00:05:54.840 the two images. 00:05:54.840 --> 00:05:59.860 It will use the data on where scratches and dust are to remove affected areas of the image. 00:05:59.860 --> 00:06:03.900 It will then fill these areas in using information from what’s around them, and it usually 00:06:03.900 --> 00:06:05.150 does a bang-up job. 00:06:05.150 --> 00:06:08.110 I am always impressed with how well this works. 00:06:08.110 --> 00:06:12.270 Now of course it’s not perfect, there are some artifacts here that it didn’t completely 00:06:12.270 --> 00:06:17.830 remove, but what’s remarkable is that in most places, it appears to be completely flawless. 00:06:17.830 --> 00:06:22.610 The algorithm responsible for deciding what to fill in the empty space with is so good 00:06:22.610 --> 00:06:25.480 that it mimics the grain structure of the film itself. 00:06:25.480 --> 00:06:29.240 You would have no idea that this area has been retouched. 00:06:29.240 --> 00:06:33.551 Pixel-peeping shows that Digital ICE had almost no effect on the clarity of the image, and 00:06:33.551 --> 00:06:36.780 appears to have done nothing except remove flaws. 00:06:36.780 --> 00:06:37.780 Here are some slides. 00:06:37.780 --> 00:06:42.320 I shot some rolls of slide film, too, which looked like this without dusting. 00:06:42.320 --> 00:06:46.500 Observe how well digital ICE is able to take care of really dusty slides. 00:06:57.700 --> 00:06:59.440 I’d like to take another brief side-note. 00:06:59.450 --> 00:07:02.530 There’s something magical about slide film to me. 00:07:02.530 --> 00:07:04.500 This IS the photograph. 00:07:04.500 --> 00:07:08.850 A slide is just a piece of the very film that was in your camera mounted in cardboard or 00:07:08.850 --> 00:07:14.200 plastic so you and slide projectors can interact with it easily and without touching it. 00:07:14.200 --> 00:07:18.610 Unlike today where cameras save digital files of ones and zeros which you need a device 00:07:18.610 --> 00:07:22.970 to be able to interpret, slide film becomes a literal photograph. 00:07:22.970 --> 00:07:26.770 The light that came through the lens and landed on the film behind it looked nearly exactly 00:07:26.770 --> 00:07:27.770 like this. 00:07:27.770 --> 00:07:32.130 It’s just really cool to me to be holding a physical object that captured light and 00:07:32.130 --> 00:07:35.940 through processing became an actual copy of that light. 00:07:35.940 --> 00:07:40.110 Anyway, here is where Digital ICE can sometimes be detrimental. 00:07:40.110 --> 00:07:44.630 Images with lots of detail tend to obscure dust on their own, particularly with slides 00:07:44.630 --> 00:07:49.080 as the dark dust is harder to notice than the bright white dust from negatives. 00:07:49.080 --> 00:07:53.630 Sometimes it really is better to not use digital ICE, as complex images can be too much for 00:07:53.630 --> 00:07:55.050 the algorithm. 00:07:55.050 --> 00:07:59.130 This image in particular seemed relatively defect-free, but Digital ICE saw a bit of 00:07:59.130 --> 00:08:03.460 dust here, and didn’t really know how to re-fill in the detail of the sign. 00:08:03.460 --> 00:08:06.680 To be honest, it would have been better to leave this alone, though it is admittedly 00:08:06.680 --> 00:08:08.520 a tiny portion of the image. 00:08:08.520 --> 00:08:10.120 Hey look, Kodak! 00:08:10.120 --> 00:08:14.419 Same with this photograph, though it was very dusty, the cacophonous nature of the leaves 00:08:14.419 --> 00:08:18.630 and organic material on the ground made it hard to notice, and this bit of dust shows 00:08:18.630 --> 00:08:20.110 how things can go wrong. 00:08:20.110 --> 00:08:24.320 It’s not hard to see once you know it’s there, but it’s easy to ignore. 00:08:24.320 --> 00:08:30.680 Digital ICE certainly saw it, and its imperfect correction ended up making it much more noticeable. 00:08:30.680 --> 00:08:35.200 Digital ICE tends to do better in areas that are low-detail such as the sky or on clothing, 00:08:35.200 --> 00:08:38.650 as the algorithm can rely on adjacent areas to fill in the defect. 00:08:38.650 --> 00:08:42.890 In these cases it merely has to mimic the grain structure and color of the areas next 00:08:42.890 --> 00:08:43.890 to the defect. 00:08:43.890 --> 00:08:48.690 If any sort of pattern or high level of detail surrounds the problem area, it can in some 00:08:48.690 --> 00:08:53.320 cases end up making the defect more noticeable as it fails to convincingly fill it in. 00:08:53.320 --> 00:08:56.060 Digital ICE also has a few technical limitations. 00:08:56.060 --> 00:09:00.570 It can only be used on color negative film or E-6 process slides, which includes most 00:09:00.570 --> 00:09:04.220 color slides such as Ektachrome and Fuji Velvia. 00:09:04.220 --> 00:09:08.360 Conventional black and white film doesn’t work because with this film, the image is 00:09:08.360 --> 00:09:10.690 composed of silver molecules. 00:09:10.690 --> 00:09:12.960 These molecules block light from passing through, and that's how 00:09:12.960 --> 00:09:15.960 it creates the dark areas of the film. 00:09:15.960 --> 00:09:20.131 The silver will also block the infrared light on the second scan, so when you use Digital 00:09:20.131 --> 00:09:24.130 ICE, the scanner will see the exact same image from both passes. 00:09:24.130 --> 00:09:28.170 It’s smart enough to realize there’s nothing it can do, and doesn’t bother trying, 00:09:28.170 --> 00:09:30.230 so you just don’t get any help. 00:09:30.230 --> 00:09:34.500 However, C41 process black and white film, which can be identified by its orange film 00:09:34.500 --> 00:09:39.080 base rather than the gray of true black-and-white film, can use Digital ICE. 00:09:39.080 --> 00:09:43.060 This type of black-and-white film was mainly sold in stores as a novelty film, designed 00:09:43.060 --> 00:09:47.660 to be compatible with the standard C41 color development process. 00:09:47.660 --> 00:09:52.690 Images on this film are composed of dyes like in normal color film, and these dyes are transparent 00:09:52.690 --> 00:09:53.690 to infrared. 00:09:53.690 --> 00:09:56.210 So, Digital ICE still works with this film. 00:09:56.210 --> 00:10:01.610 Kodachrome slide film also doesn’t behave too well with Digital ICE, though it can depending 00:10:01.610 --> 00:10:02.930 on the slide. 00:10:02.930 --> 00:10:07.620 In this case, the cyan dye in the film can absorb some of the infrared light, which occasionally 00:10:07.620 --> 00:10:09.540 becomes detected as a defect. 00:10:09.540 --> 00:10:14.150 The scanner might try to remove things it shouldn’t, and the results can be weird. 00:10:14.150 --> 00:10:18.210 I don’t have any Kodachrome slides to test this on, as although Simon and Garfunkel tried 00:10:18.210 --> 00:10:22.310 their best to convince her otherwise, mama took my Kodachrome away in 2010. 00:10:22.310 --> 00:10:25.680 However, google Kodachrome and Digital ICE and you’ll see some good examples. 00:10:25.680 --> 00:10:28.690 I’ll link a particularly good one in the description. 00:10:28.690 --> 00:10:32.500 The biggest drawback to using Digital ICE is that it more than doubles the time it takes 00:10:32.500 --> 00:10:36.160 to scan film, an already a tediously slow process. 00:10:36.160 --> 00:10:40.000 Which brings me back to why this scanner is such an upgrade from my old one. 00:10:40.000 --> 00:10:45.050 The CCFL backlight of my old scanner required about a one minute warmup between each successive 00:10:45.050 --> 00:10:50.330 scan when using Digital Ice, as the lamp would cool back down during the infrared scan. 00:10:50.330 --> 00:10:52.690 This made using it a real chore. 00:10:52.690 --> 00:10:57.260 A fully loaded film holder would take another 12 minutes on top of the more than doubled 00:10:57.260 --> 00:11:00.540 scan time needed for the second scan and image processing. 00:11:00.540 --> 00:11:04.470 This could easily make scanning 12 negatives take over a half hour. 00:11:04.470 --> 00:11:08.900 It still takes this scanner about twice as long to perform a scan with Digital ICE, but 00:11:08.900 --> 00:11:10.770 I think it works wonders. 00:11:10.770 --> 00:11:14.490 Like I said before, some people refuse to use it, and it is true that for slides and 00:11:14.490 --> 00:11:18.640 negatives that aren’t damaged, a duster spray can is capable of producing a nearly 00:11:18.640 --> 00:11:19.730 flawless image. 00:11:19.730 --> 00:11:23.900 And it’s also true that Digital ICE can occasionally produce weird glitches of its 00:11:23.900 --> 00:11:27.430 own, like how it made this piece of dust look worse after correction. 00:11:27.430 --> 00:11:32.800 But I still think being able to go from this, to this, without any sort of input is just 00:11:32.800 --> 00:11:33.970 amazing. 00:11:33.970 --> 00:11:37.870 When it works, which it usually does, it’s nothing short of awesome. 00:11:37.870 --> 00:11:41.530 Thanks for watching, I hope you learned something interesting in this Tech Exploration (though 00:11:41.530 --> 00:11:44.870 if you’re a photography buff, you probably didn’t). 00:11:44.870 --> 00:11:48.060 If you’re new to this channel and you liked what you saw, why not subscribe? 00:11:48.060 --> 00:11:49.300 There’s more on the way. 00:11:49.300 --> 00:11:53.460 I’d also like to thank all of my Patreon supporters for making this channel possible. 00:11:53.460 --> 00:11:55.770 Your contributions are appreciated each and every day. 00:11:55.770 --> 00:11:57.620 I’ll see you next time.