A conversation with Zebra Dry Plates founder Nejc about rebuilding Autochrome, turning craftsmanship into machinery, and whether a century-old colour process can become something photographers can use again.

In 2017, Nejc was a student with a slightly unreasonable idea: make Autochrome.

Invented by the Lumière brothers and introduced commercially in 1907, Autochrome was the first widely successful colour photographic process. Its colour screen was made from microscopic grains of dyed potato starch, with a photographic emulsion coated on top. Expose the plate, process it, hold the finished transparency to the light, and those tiny grains reconstruct colour.

Beautifully simple in description, but absurdly difficult in practice.

Nejc quickly discovered that before he could make Autochrome, he first had to learn how to make photographic emulsion. That prerequisite became Zebra Dry Plates, and what was supposed to be preparation for an experiment eventually became years of emulsion making, tens of thousands of hand-coated glass plates, machinery, chemistry, sensitometry, packaging, shipping and, eventually, a factory.

Autochrome, meanwhile, remained unfinished, until this April, when there was finally an image on the plate.

We sat down with Nejc to talk about the nine-year journey back to Autochrome, why potatoes are considerably more complicated photographic materials than they sound, and the much larger question now facing us: if we can make Autochrome, can we build the machinery necessary to make it available to everyone else?

Was Zebra really a nine-year detour back to Autochrome?

You picked Autochrome as a student in 2017, and learning to make emulsion was only supposed to be the way in. Nine years later, the emulsion is a factory and the Autochrome finally arrived in April. From outside, Zebra looks like a very long detour back to an unfinished student project. Did the detour become the real work, or has the workshop always been the means of getting back to Autochrome?

Nejc: Calling Zebra a nine-year detour back to Autochrome is actually quite accurate, although I didn’t understand it that way while it was happening.

Autochrome was the reason I started making photographic emulsions in the first place. In 2017 I was a student and the idea was essentially: I want to make Autochrome, therefore I need to learn how photographic emulsion works.

That sounds wonderfully straightforward when compressed into one sentence. In reality, the prerequisite became a subject large enough to consume years.

At some point the detour became the work. Making an emulsion once is one problem. Making it reliably is another. Making hundreds or thousands of plates which behave the same way is another again. Then you have glass, coating, drying, packaging, sensitometry, machinery, chemistry, customers, shipping. Eventually you look around and what began as preparation for an experiment has become a factory.

But Autochrome never really disappeared. It sat in the background as a sort of unfinished sentence. There were always more immediate problems to solve, and every one of those problems was also giving me another piece of the knowledge I would eventually need to return to it.

When the plate finally worked in April, the strange thing was that I was simultaneously much closer to and much further from the person who imagined it in 2017. The student version of me imagined the photograph. He didn’t imagine the nine years of infrastructure required to make the photograph possible.

And no, it wasn’t exactly the thing I imagined then. It couldn’t be. In 2017 Autochrome was almost mythical to me. Once you have actually made one, it stops being mythology and becomes materials: starch, dyes, carbon black, gelatin, silver halides, glass.

In one sense that demystifies it. In another, I find it much more extraordinary now because I understand how absurdly difficult it is that all of those things work together at all.

What does a hand know that a machine doesn’t?

Before building your coating machine, you had hand-poured something like a hundred thousand plates and refined the syringe-and-tilt method over roughly sixty thousand of them. Then you built a two-tonne machine. Teaching it meant turning decisions that lived in your wrist into numbers: speed, viscosity, temperature, flow. What did the hand know that was hardest to write down as a parameter?

Nejc: The hardest thing to translate was not a particular movement of the wrist. It was a correction.

When you coat enough plates by hand, you are never really repeating exactly the same movement. You are continuously compensating without consciously describing what you are compensating for.

The emulsion feels slightly different today. The glass is a different temperature. The bead is moving too quickly on one side. You tilted a fraction too far. You see the coating front behaving differently and your hand changes what it is doing before you have formulated the problem in words.

That is the knowledge hidden in the wrist.

A machine cannot have that kind of intuition unless you first identify what the intuition is responding to. It forces you to separate one apparently simple act, “coat the plate”, into viscosity, temperature, volume, acceleration, velocity, angle, timing and so on. Then you discover how many of those variables interact.

That was probably the most interesting part of building the machine. I had spent years training myself to compensate for an imperfect process. With the machine, instead of getting better at compensating, I could start removing the reasons compensation was necessary.

There are still things that are difficult to translate. A human is exceptionally good at noticing that something simply looks or feels wrong. But the answer isn’t to romanticise that. If I can identify what I am noticing, I would rather turn it into a measurable parameter.

The machine is, in some ways, a record of everything hand coating taught me.

And I should probably defend my poor machine on one point. I didn’t build two tonnes of machinery because I never wanted to pour another plate. I still enjoy hand coating. I built it because after tens of thousands of plates I no longer thought variability was something the customer should have to pay for just because it proved a human had been involved.

Does mechanising a handmade process destroy something important?

That raises a slightly uncomfortable question. Hand-pouring is exactly what a lot of people imagine they’re buying from a small photographic manufacturer. A machine-made plate can be thinner, more even and more consistent. Better by almost every measurable standard. But does the customer lose something real?

Nejc: This is where I think “handmade” can become a slightly dangerous word.

If somebody buys a handmade chair, perhaps the marks of the maker are part of the object. With a photographic plate, the coating is not really the final object. It is the material from which the photographer is going to make their object.

A bubble isn’t my signature, dust isn’t craftsmanship, an uneven coating isn’t evidence that somebody cared more.

I don’t think the customer loses anything photographically by having a thinner, cleaner and more uniform emulsion. They gain predictability.

If there is an irregularity in the final photograph, I would much rather it came from the photographer, the lens, the subject, the development or chance than from the fact that my wrist moved differently on plate number 38,412.

What may be lost is a certain idea of handmade production. I understand that. There is something attractive about knowing somebody stood in a darkroom and poured your individual plate. I did exactly that for years, and it is part of Zebra’s history. But the machine hasn’t removed the hand from the process. We still make the emulsion. We prepare the glass. We operate the equipment. We inspect the plates. We test them.

The machine only performs one operation better and more consistently than I can. If the highest expression of craftsmanship is deliberately preserving defects that you already know how to remove, then I don’t find that definition of craftsmanship very useful.

And then there are the imperfect plates

You’ve also started selling Test Plates with minor defects at a lower price. So the flaw itself has now become a line in the price list. Where is the boundary between a defect worth eliminating and one worth selling?

Nejc: The distinction for Test Plates is quite simple for me: does the defect prevent the plate from doing useful photographic work?

There are defects that compromise the material, sensitometric problems, serious coating failures, contamination, anything that makes exposure unpredictable. Those don’t become cheaper products. They’re a waste.

Then there are cosmetic or local defects where most of the plate is perfectly usable. Historically I was extremely reluctant to sell those because I wanted every Zebra box to represent the standard I was aiming for.

But throwing away a photographic plate that somebody could make a perfectly good image with is also difficult to justify, particularly when the customer knows exactly what they are buying and pays accordingly.

So Test Plates are not a lower quality target. They are the consequence of having a quality target. And yes, I fully expect somebody to become nostalgic for the old irregularities. Photography is very good at becoming nostalgic for technical limitations approximately five minutes after we solve them.

Maybe in twenty years I’ll have to build a machine that accurately reproduces the defects of my hand-coated plates.

ISO 2 changes more than the exposure time

Your plates are around ISO 2, roughly seven or eight stops slower than ordinary film. At that point slowness isn’t just an inconvenience. It changes what can be photographed. What becomes impossible?

Nejc: ISO 2 removes movement from your vocabulary very quickly.

There are photographs you simply cannot make in the conventional sense. Fast action, spontaneous gestures, many kinds of street photography, anything where the decisive moment lasts a fraction of a second. The subject will be gone long before the exposure is finished.

Even wind becomes a photographic decision. Leaves, water, clothing and hair behave differently when exposure is measured in seconds rather than fractions of seconds.

Portraiture is particularly interesting because the exposure becomes a collaboration. You cannot steal the expression in quite the same way. The person has to participate in making the photograph. They have to remain there. Even when they try to remain perfectly still, they don’t, so time becomes visible in their faces. I like that.

But I don’t think slowness should be fetishised either. ISO 2 is partly an aesthetic characteristic and partly a limitation of the material. We shouldn’t pretend every limitation is profound simply because it is old.

The panchromatic work isn’t primarily an attempt to make Zebra into modern high-speed film. Panchromatic sensitivity is important because it changes how colours translate into tone and, of course, it is essential for colour processes such as Autochrome.

Speed can improve as the emulsions improve, and I’m interested in seeing how far we can take that. But I have no ambition to make a dry plate behave exactly like a modern film. If that is what somebody needs, modern film already exists and is exceptionally good at it.

The constraints are part of what makes plates interesting. The important distinction for me is between a constraint that creates a different photographic language and a defect that merely gets in the way.

Now turn off the lights

Autochrome needs a panchromatic emulsion, and panchromatic material tolerates no safelight. So one of the most beautiful colour processes ever invented has to be coated in complete darkness. How do you judge an even coat when you cannot actually look at it?

Nejc: Complete darkness changes the psychology of coating more than I expected.

With orthochromatic material, even under a very dim safelight, vision gives you reassurance. You can inspect the coating front, see where the emulsion is moving, see whether a corner has been covered.

With panchromatic emulsion that feedback disappears.

So you replace sight with repetition and physical information. You know the amount of emulsion you’ve dispensed. You know the position and angle of the plate. You feel how the liquid moves through the plate and through your hands. Timing becomes much more important. You learn not to interrupt a movement simply because you can’t visually confirm it.

In a strange way, making panchromatic plates manually asks you to trust everything you learned while you could see. And then you switch the lights on later and discover whether your confidence was justified.

That is also one of the places where machinery becomes particularly valuable. A machine doesn’t become less accurate because somebody turned the lights off. Once the parameters are established, darkness is irrelevant to it.

Which brings us to potatoes

Autochrome has one of those descriptions that sounds almost too wonderful to be true: microscopic grains of dyed potato starch, roughly ten to fifteen microns across, forming the colour screen. What does “ten to fifteen microns” actually mean when you’re trying to manufacture it?

Nejc: “Potato starch” makes Autochrome sound charmingly domestic. The reality is closer to particle engineering. You need starch grains within a very particular size range because every grain is effectively part of the colour screen. Too large and the structure becomes increasingly visible, and you sacrifice resolution. Too fine and you introduce other problems in processing and handling.

So getting to roughly 10 to 15 microns means separating and grading an enormous population of particles rather than finding some magical variety of microscopic potato. Then those grains have to be divided, dyed into the required colour components, brought back together in controlled proportions and distributed as densely and randomly as possible. The gaps have to be dealt with, the layer has to be flattened, and all of this has to remain compatible with the photographic emulsion sitting above it.

That is one of the things I love about Autochrome. Before you have made any silver image at all, you have already manufactured a microscopic colour filter out of potatoes.

Could a different batch of potatoes actually change the colour of the photograph?

Nejc: The properties of the starch matter, although I wouldn’t describe it as simply “different potatoes equal different coloured photographs.”

The colour is primarily determined by the dyes, their concentration, the proportions of the coloured grains, the spectral response of the emulsion and the processing. But particle size, shape, transparency and the way the starch accepts the dye can all affect the behaviour of the screen So a change in raw material can absolutely propagate into the final image. Whether the potato gets the blame personally is another question.

There is an image on the plate. But it isn’t finished yet.

The first complete plate produced a colour negative.

That matters because historically an Autochrome is a positive transparency. You hold it to the light and the image itself glows through its microscopic colour screen. Reliable reversal of hand-coated plates remains another problem to solve.

So have you made an Autochrome if the result isn’t yet a positive? Is it a failed reconstruction, or are we already making something new?

Nejc: I wouldn’t call the negative a failed Autochrome, but it isn’t an Autochrome in the strict historical sense either. The historical Autochrome is a colour transparency. Light passes through the developed silver image and the microscopic colour screen, so what you see is a positive colour image. Reversal isn’t really an optional final step. It’s fundamental to what an Autochrome is.

What we demonstrated with the negative was slightly different, but at that stage it was actually more important to me that we could see the screen and the emulsion working together photographically. The colour information was there and the plate was producing an image. After nine years of working on all these separate pieces, that was the first time we knew they could actually work together. Reversal is another problem, and that’s what we’re working on now.

I’m also not particularly interested in claiming that we’ve resurrected the Lumières’ Autochrome exactly as it existed in 1907. Even if you reproduce the published formula perfectly, you’re still working with different gelatin, different dyes, different glass, different chemicals and different equipment. There’s always going to be some interpretation when you’re trying to reconstruct a historical process. What interests me more is understanding the principles well enough that we can make the process work again, rather than simply trying to reproduce an artefact.

So the negative is evidence. Nine years ago, I was a student with what seemed like a pretty ridiculous idea: that I might be able to make a colour photograph from glass, silver and microscopic grains of dyed potato starch. Now there is an image on the plate. We just need to make it positive.

The next problem is scale

This is where the Autochrome project arrives at a very different point. For nine years, the central question was whether we could make it work at all. Now we have to ask whether we can make it repeatable.

Making one plate by hand is an experiment. Making Autochrome available as a photographic material is manufacturing, and the process is considerably more complicated than coating an ordinary dry plate.

The starch has to be selected and graded to an extremely narrow particle size. It has to be separated into colour components, dyed consistently, recombined in controlled proportions and distributed into a dense microscopic screen. That layer has to be processed and flattened before a panchromatic photographic emulsion can be applied over it.

Because that emulsion is panchromatic, parts of the production process must happen in complete darkness. It is possible to do extraordinary things by hand. Zebra exists because an unreasonable amount of photographic material was made that way. But if Autochrome is going to become something other photographers can actually buy, expose and use, we need to do what we eventually had to do with dry plates: take knowledge that currently lives in hands, experiments and individual procedures and turn it into specialised machinery that can repeat those procedures accurately.

That machinery does not exist on a shelf waiting for us to order it. We will have to design and build much of it ourselves. And that requires a considerable amount of funding.

Should we crowdfund Autochrome?

This is something we would genuinely like to ask the photographic community.

We could continue developing the process internally and move forward as resources allow. That is the slower route, but it keeps development entirely within Zebra. The other possibility is crowdfunding.

A crowdfunding campaign could allow us to build the specialised machinery needed to mechanise the Autochrome production process much sooner. More importantly, it could tell us whether there is a community of photographers who want this process badly enough to help us bring it back into practical use.

We are not interested in crowdfunding an idea that exists only as a rendering or a promise. There is already an image on the plate. The question now is whether we should build the machines that could put plates like it into other photographers’ hands.

Nine years later

There is something satisfying about the fact that the biggest obstacle to Autochrome may turn out to be the same problem that came from our original attempt to recreate it: how do you take something you can make once, with enormous care, and make it reliably?

That question is what took us from photographic emulsion experiments to Zebra Dry Plates. From syringes and hand-tilted glass to a two-tonne coating machine. From relying on the movement of a wrist to measuring viscosity, temperature, acceleration, flow and timing.

And now we are facing that transition again, only this time the material on the other side is colour.

The goal isn’t simply to reproduce an object from 1907, or to remove every strange limitation that makes Autochrome what it is. What interests us is somewhere in between: understanding the process well enough that it can become a living photographic material again.

Nine years ago, that was just an idea. Today, there is an image on the plate. Now we have to figure out how to make the next thousand.