Transferring vintage Kodacolor lenticular film

Todd Ruel · July 9, 2024

I recently asked my Galileo Digital rep if Lasergraphics had any solutions for successfully digitizing vintage Kodacolor film made between 1928-1935.

I got back a word salad explanation of Kodacolor and a suggestion to "hire a software engineer to write code to convert the lenticle stripes to color (or gray-scale if the scan is B&W). To make the output look good will presumably be difficult and expensive." (Sure, I'll just get right on that.)

As it turns out, Tom Aschenbach of Colorlab has been working on this for years and has written his own software. I found this interesting link on LinkedIn: https://www.linkedin.com/feed/update/urn:li:activity:7211366878466719744/. It's his third version of his own software for Kodacolor lenticular film.

As I understand it, Tom was also instrumental in the development of AEO Light, the optical sound extraction app available for free on the interwebs.

Robert Houllahan · July 10, 2024

I have scanned some Kodacolor jobs and sent them to Tom at Colorlab for him to process, his GPU based software works and the results are certainly a mixed bag.

Kodacolor was a terrible color process and was made far worse by years of being squished so the lenticular stripes became flat.

At best it is a low res color image with lines and at worst it is a low res B&W image with lines when the processing does not quite see the stripes, I am sure Tom has improved it over time as it has been a few years since I have seen any7 Kodacolor come into the lab.

Perry Paolantonio · July 10, 2024

writing software to do this is something we've messed with as well. It's not difficult in theory, but the biggest problem is that due to shrinkage and warping, the vertical lenticules get out of whack and you wind up with rainbow moire effects. To do it right involves a fair bit of image processing to compensate for all of that first, before applying color to the three channels. It's doable, but a lot of work.

We hired someone a while back to write some code for us to do this, and it kind of worked but only if the film was absolutely pristine and flat, which is almost never the case with 100 year old film...

Todd Ruel · July 11, 2024

14 hours ago, Perry Paolantonio said:

It's doable, but a lot of work.

I'm guessing that's why Lasergraphics has not created some sort of $11K software license for this like they did with 2-Flash HDR. Too much work for so little content. Also, results are unpredictable, because the condition of the source material is unpredictable.

Nevertheless, it's kinda breathtaking to see the 1933 Chicago World's Fair in color. Or the Royal Family from 1930. Or Charles Lindbergh in color.

I guess I'm shipping that stuff off to Colorlab!

Daniel D. Teoli Jr. · July 13, 2024

Maybe AI will fix it!

Joerg Polzfusz · July 13, 2024

Hi! Can’t you simply add an overlay of colored stripes and then squeeze the result so that the three neighboring stripes overlap? Or is some information missing as the film‘s surface isn’t flat?

Edited July 13, 2024 by Joerg Polzfusz

Joerg Polzfusz · July 14, 2024

On 7/10/2024 at 3:57 PM, Perry Paolantonio said:

…it kind of worked but only if the film was absolutely pristine and flat, which is almost never the case with 100 year old film...

Okay, then please ignore my idea from the previous post. 😉

However one question remains: Is a scanner able to properly work with the non-flat surface of the film at all? Or will this cause stray-lights/flares/…?

image.jpeg.f91511ea84752d570c771208c5881942.jpeg

Edited July 14, 2024 by Joerg Polzfusz

Perry Paolantonio · July 15, 2024

On 7/14/2024 at 3:16 AM, Joerg Polzfusz said:

However one question remains: Is a scanner able to properly work with the non-flat surface of the film at all? Or will this cause stray-lights/flares/…?

The scanner (at least ours) can handle warped film, and there aren't any real side effects except for some slight blurring in extreme cases. There are options to install a modified gate, which has a pressure plate to hold the film flat. I think the one from Lasergraphics is too expensive so I've been designing my own. Yet another project...

The issue here with lenticular film is that there are 400 or so lenses baked into the film itself, and those are tiny. Any slight variation causes issues when applying the color to the R/G/B channels that correspond to those lenticules. So any software that deals with this first has to make sure all those lines are perfectly straight. That's doable, but it's a lot of pre-processing, and in cases where the film is really dense it can be hard, if not impossible, to detect the edges of those lines. So you have to make some guesses about where they should be based on what you know of what you can actually see. That's where things get tricky.

Robert Houllahan · July 16, 2024

On 7/11/2024 at 12:20 AM, Todd Ruel said:

I'm guessing that's why Lasergraphics has not created some sort of $11K software license for this like they did with 2-Flash HDR. Too much work for so little content.

The 2-flash is useful for allot of things and mostly gets sensor noise down and more detail in dense film so a reasonable cost for that feature and LaserGraphics has a staff of engineers and support people to feed.

Tommy at Colorlab has a background in software and Colorlab probably sees more of that Kodacolor film than anyone with their proximity to the national archive so it was likely worth it for him to develop it.

It is pretty problematic as Perry said with defining the little squashed lenses from so long ago and there is so little of this kind of film in circulation.

Dan Baxter · July 16, 2024

On 7/11/2024 at 2:20 PM, Todd Ruel said:

I'm guessing that's why Lasergraphics has not created some sort of $11K software license for this like they did with 2-Flash HDR. Too much work for so little content. Also, results are unpredictable, because the condition of the source material is unpredictable.

It's not really their job to offer software that does things in the post-processing step as the scanner manufacturer. A $20,000 software licence to make this?

Not sure who would buy it.

It's good you identified the person working on it, that's where that work belongs. In the hands of someone capable taking it on as additional work for the sake of preservation.

John Rizzo · July 16, 2024

Tommy from Color Lab wrote custom software for lenticular conversion:

https://www.linkedin.com/feed/update/urn:li:activity:7217533019249635328/

Todd Ruel · July 17, 2024

17 hours ago, Dan Baxter said:

It's not really their job to offer software that does things in the post-processing step as the scanner manufacturer.

That doesn't make any sense to me. I could consider image stabilization to be a post-processing step. I could consider audio extraction to be a post-processing step. Indeed, both of those are post-processing steps for other film scanners like Filmfabriek and Kinetta.

Why should processing Kodacolor inline with the capture process be any different? Nevertheless, Lasergraphics has chosen to include those other features with their scanners, and those features are real differentiating factors.

My main point was to ask: why can't the Cadillac (or insert your favorite brand of luxury car here) of film scanners become a McLaren by adding a super exclusive feature like processing vintage Kodacolor? The answer is still the same: too little content to justify the expense of development.

It's the perfect challenge for someone like Tommy Aschenbach.

Dan Baxter · July 21, 2024

On 7/17/2024 at 2:57 PM, Todd Ruel said:

That doesn't make any sense to me. I could consider image stabilization to be a post-processing step. I could consider audio extraction to be a post-processing step. Indeed, both of those are post-processing steps for other film scanners like Filmfabriek and Kinetta.

Says who that image stabilisation belongs in post-processing? There was a trend away from the "telecine process" where the transfer goes straight to broadcast tape/finishing format. Most of the pin-registered machines don't even show the perfs in the scan/transfer either as they used the full width of the sensor for the picture. Image stabilisation in-scan is part of the optical pin registration, LG calculate how much the frame has moved between frames to correct for the next capture. The better the optical pin registration is the more stable it should be in-scan, requiring you do less in post.

The number of different audio formats (whether hardware or software based) also gives them an advantage over competitors. Software audio extraction on a LG is better quality than AEO Light as you may have discovered if you've done some tests.

On 7/17/2024 at 2:57 PM, Todd Ruel said:

My main point was to ask: why can't the Cadillac (or insert your favorite brand of luxury car here) of film scanners become a McLaren by adding a super exclusive feature like processing vintage Kodacolor? The answer is still the same: too little content to justify the expense of development.

They're not a film restoration software company, they're a scanner manufacturer. Arri and DFT also don't write Kodacolor decoding software. Doing it the way you're thinking would bake it in to the scan, meaning you can't go back to the start and start-over if you discover a better process in the future for the decoding.

Perry Paolantonio · July 22, 2024

On 7/17/2024 at 12:57 AM, Todd Ruel said:

That doesn't make any sense to me. I could consider image stabilization to be a post-processing step. I could consider audio extraction to be a post-processing step. Indeed, both of those are post-processing steps for other film scanners like Filmfabriek and Kinetta.

I consider the Lasergraphics scanners to be stable, in terms of image registration. But I don't really consider them to be "doing stabilization," which often has other meanings. We get a lot of clients who apply post stabilization to their scans because they're unhappy about the shakiness of the image. But that shakiness is baked into the picture. The job of the scanner, IMO, is to reproduce the image digitally as faithfully as possible, and then to do fine tuning later. In the case of something shot on a very stable camera, the scan is often all you need because the picture within the frame is stable to begin with. But for stuff shot on cheap cameras like the K3, it's not uncommon to need more work in post. I don't think that really belongs in the scanner.

Kind of the same as color grading. We will offer a one-light scan if someone really wants it but we don't do scene by scene in the scanner because it's just not the way modern scanning workflows are designed. With tools like Resolve available for free, and more capable than color correction systems from 15-20 years ago, it makes no sense to break down a film scene by scene in the scanner and apply a grade - that's a lot of wasted time that could be used scanning other film. The color grading tools in Resolve are way better than what you get in any scanner software anyway, so that's the place to do that.

So I feel the same about things like lenticular film. I would rather get the best digital reproduction of the film I can, then do the color reconstruction in post, vs permanently baking the color work into the scan. That leaves open the possibility to revisit it later with potentially better color restoration tools too, without having to scan. And on film that's 100 years old, the possibility of not being able to re-scan it 10 years from now is very real.

Todd Ruel · August 1, 2024

On 7/21/2024 at 7:29 AM, Dan Baxter said:

Says who that image stabilisation belongs in post-processing?

I don't think of it as belonging anywhere. I'm thinking from the bottom up. The Wolverine and Filmfabriek scanners, for instance, don't have image stabilization. You have to do that in post-production. So I consider it a post-production process that others, like Kinetta and Lasergraphics, do inline. This is a semantic argument. You say tomato. I say to-mah-to.

On 7/21/2024 at 7:29 AM, Dan Baxter said:

Doing it the way you're thinking would bake it in to the scan, meaning you can't go back to the start and start-over if you discover a better process in the future for the decoding.

On 7/22/2024 at 9:29 AM, Perry Paolantonio said:

I would rather get the best digital reproduction of the film I can, then do the color reconstruction in post, vs permanently baking the color work into the scan.

Two things:

1) All of these inline features have an OFF button. You don't have to turn them on. Why worry about baking a look into a scan when nobody is forcing you to do that?

2) I don't do a tremendous amount of archival work. I recently scanned 16mm football games from 1970 for a customer. I'm not too worried about revisiting this material 10 years from now. It's not that precious. I always advise my customers not to throw away their films after scanning them, because scanning technology has greatly improved in the last 15 years. It might do so in another fifteen. But I have to take a more practical point of view about this material. Not everything can or will be considered archival.

On 7/22/2024 at 9:29 AM, Perry Paolantonio said:

I consider the Lasergraphics scanners to be stable, in terms of image registration. But I don't really consider them to be "doing stabilization,"

If the scanner is, in any way, stabilizing the image like Lasergraphics does through software-based optical pin registration, I consider that to be "doing stabilization." Again, I'm approaching this from the bottom up. Lesser scanners don't do this inline. Lasergraphics does, and I think that's fantastic. That's one of the reasons I bought one.

Once again: will any of these companies offer a feature like interpreting vintage Kodacolor as an inline scanning feature? I seriously doubt it for super insightful reasons you've all listed above.

Perry Paolantonio · August 1, 2024

8 hours ago, Todd Ruel said:

If the scanner is, in any way, stabilizing the image like Lasergraphics does through software-based optical pin registration, I consider that to be "doing stabilization."

I understand what you're saying, but I guess my point was that I consider that such a basic feature it should simply be considered part of the scanner. I mean, a pin-registered scanner is going to get you basically the same result (at least for non-shrunken film), but it's doing it mechanically. It's just assumed that your image should be stable. That's part of why I don't consider the Filmfabriek, MovieStuff, etc to be serious machines.

8 hours ago, Todd Ruel said:

I don't do a tremendous amount of archival work. I recently scanned 16mm football games from 1970 for a customer. I'm not too worried about revisiting this material 10 years from now. It's not that precious.

Sure, but we're talking about kodacolor, which existed for a few years, 100 years ago. It is, by definition, archival film. The odds of it surviving another 20 years without serious shrinkage issues are slimmer now than they were 20 years ago. So the right approach is to get the data captured as best you can and then deal with the color in software. You can apply the same logic from your statement "I always advise my customers not to throw away their films after scanning them, because scanning technology has greatly improved in the last 15 years" to a software based solution, and given that the film will likely not be in the same condition 15-20 years from now, because the software to do this may very well improve over time.