WARNING:  PAGE UNDER CONSTRUCTION.  A lot of the information may be missing or just plain wrong.  I'm sure I made a ton of spelling mistakes too.  Also the mobile version is SUPER broken right now, this page is best viewed on a desktop.

Manufacturing the Screen Plate


The screen plate is the most unique and recognizable part of the autochrome process, but not necessarily the most difficult to make.  Three portions of ordinary potato starch, dyed orange-red, green and violet, are mixed together to form a neutral reddish-grey, and then dusted onto a glass plate.  The glass has been coated in a thin, optically clear latex-based varnish, and the potato starch sticks readily to it.  The starch is compressed with a roller.  This compression is necessary, as the amount of light that can pass through a flattened grain is vastly more than unattended, increasing the overall speed of the plate, as well as overall brightness.  Charcoal is brushed on to fill in any open space in between the grains, and the whole thing is covered in a protective coating called the "second varnish".  

The final product is a very fine, random, color separation matrix, to which we would apply a panchromatic black and white emulsion.  There is a ton of work that goes into each and every plate.  Luckily, with a correct choice of second varnish, the screen can be reused indefinitely if an exposure doesn't work out. 


Part 1 - Preparing the Starch (Optional)


Ordinary potato starch can be used here.  For most of my work, I've always used "Bob's Red Mill" brand potato starch.  

Original autochromes separated the smallest starch grains from the larger ones via levigation -- they mixed it with water, where the largest particles would fall out of suspension first.  After a period of time, the water containing the smaller particles would be transferred to another container where they could settle out.  

This separation is not strictly necessary, but will make the final resulting screen less grainy.  All the plates featured in this post all use unsorted starch -- I have not yet made a screen plate with the finer stuff.  I have sorted and dyed some of it, so I figured I'd include some notes here.

To the right is a picture of my simple levigation apparatus.  It consists of two 5 gallon buckets.  The upper bucket has a ball valve mounted a few inches up from the bottom.  To begin, I add a full 1L measuring cup of potato starch to the bottom of the upper bucket, and fill the thing nearly to the top with water from a hose.  I take a long spatula and thoroughly mix the starch into the water, making sure the thick deposit of starch at the bottom has all been suspended.  The water is allowed to sit for 1 hour, and then the upper bucket's valve is opened, draining the contents into the lower bucket.  The lower bucket is removed, and the starch allowed to settle overnight.  After the starch has properly settled, the water can poured out without much loss of product, since most of it is now stuck in a thick layer on the bottom.  The sorted starch can be collected and dried.


Recovery is quite low, so obtaining a reasonable amount of starch to dye will take several times.  Usually I'll repeat the process with a second lower bucket using the same batch of starch in the upper bucket.  Recover is even lower than the first time, and I figured additional runs will yield vastly diminished returns.


Part 2 - Dyeing The Starch


Whether you chose to sort or not, the next step here is to dye the starch a few different colors.  The original dye formulas can be found in various places online, and I tried to stay as faithful as I could to them.  The red and violet dyes are more or less the same, just scaled down quite a bit.  I've found, at least with the dyes I sourced, that the green starch came out quite a bit more yellow than it should.  I increased the amount of blue dye and decreased the amount of yellow to hopefully make it slightly more green.  

Over time, I also started omitting a few other extraneous steps that were in the original Lumière operating procedure -- I do not heat the violet dye solution (my now purple equipment is a testament to the fact that it stains quite well enough as it is), and I do not add ammonia or sodium sulfate to the green starch.  The ammonia never seemed to alter the colors in any way, and the sodium sulfate tended to crystalize into large white chunks that contaminated the green starch.

Here are a few sources that I used for my dyes as a few of them can be a little tricky to obtain.


Crystal Violet (eBay)
Erythrosine (eBay)

Malachite Green (eBay)

Patent Blue V (Fast Colours)

Tartrazine (eBay)


This seems to be a source for Rose Bengal in German here (big thanks to @geoffreys_photography for finding it).


I first dissolve the dyes into 150mL of water in a 250mL beaker with magnetic mixing, and allow the dye to dissolve completely over 10-15 minutes or so.  The green and violet solutions will be so dark that they appear black, and all of them become a bit syrupy.  After I'm certain they're all dissolved, I slowly add 50 - 150g of starch into the beaker, still under magnetic stirring.  After the addition, I give it a good hour of stirring before transferring the slop to my vacuum filtration setup.  Early on I would use simple gravity filtration with a coffee filter, but vacuum filtration has several advantages -- Firstly, it removes the bulk of the dye much more quickly than via gravity (or a hand powered pump).  Second, the quick removal of the dye from the starch tends to allow for a more consistent coloring, whereas gravity filtration causes some grains to be dyed more heavily than others.  Even under a vacuum, removal of the thick dye solution can take 10 or 20 minutes -- patience is key.  Adding a bit of denatured alcohol (up to 100mL or so) can thin things out if the slop is too thick to be filtered. 


In my early trials I would toss the dye solution each time, but this proved to be very wasteful -- I now save and store the dye solution after filtration to be reused.  This helps to be able to dye multiple batches of starch with a higher degree of consistency.  


After the starch has been collected and dried, you may find that it tends to clump up a bit.  In particular, I tend to have the most trouble with the orange starch, as it forms tiny granules that stubbornly stuck together. 

To unclump the starch a bit, I first add it to a mortar and pestle and mash it up.  I then tip the contents through a filter strainer funnel I built.  A lot of larger clumps won't pass through, but these can be forced through the mesh with the pestle.  In general, this will break up the vast majority of the starch into a nice, smooth powder.  In my experience, the green and violet starches don't need too much more work than this.  If you experience more problems, finer filters can be bought cheaply off of eBay.  I personally will strain any larger chunks through the 120 micron filter, returning what doesn't pass back to the mortar and pestle to be ground up more finely.  If you're using sorted starch, I would strain everything again through a 53 micron strainer as well.  

Part 3 - The First Varnish


We can now mix the "First Varnish".  It's a thin, optically clear layer that adheres to the glass plate.  It provides a tacky surface that the starch will readily stick to when gently dusted on.  


Traditionally, the Lumières dissolved natural rubber in toluene, with a little bit of waxy damar beta-resene.  I struggled trying to source for raw latex that would dissolve, even going as far as to try out unlubricated condoms.  Nothing seemed to work out.  Fortunately, we have a very effective alternative that is stocked in nearly every local hardware store -- rubber cement!  I weighed out, dried, and weighed again a portion of rubber cement to determine roughly how much rubber it contained.  The rubber cement is then diluted with xylene (toluene will work just fine too if you have some on hand) to match the proportions of the original Lumière varnish.  You will note -- in my formula, there is no damar component.  I'm not sure why this was added to the original varnish, but I found that it had a tendency to reduce the tackiness of the plate, rather than enhancing it.  The formula I still use is as follows:

The First Varnish:

  • Elmer's Rubber Cement, 28g

  • Xylene, 200mL

Measure out the xylene into a 250mL beaker, and place on a magnetic mixer.  Weigh out the rubber cement in a disposable weigh boat.  While stirring, slowly tip the weigh boat and allow the rubber cement to flow into the beaker.  Mix for another 5 minute or so, until the solution is clear and homogeneous.  Gravity filter the solution through a coffee filter and bottle it.  This stuff will last a real long time. 

Note:  You're not going to get insta-cancer just by getting a whiff of it, but xylene/toluene fumes aren't great for you health either.  Whenever working with the varnish, whether it's mixing it up or coating glass plates, you should probably be doing it either in a fume hood or outside.  

Part 4 - Mixing and Dusting the Starch


We have our three portions of starch, now we need to mix them together correctly to form a neutral-colored screen!  One might think that this is as simple as weighing out the same amount of each color, but unfortunately it's not quite that straightforward.  The same mass of starch of each color may not correlate to the number of grains anymore, since the different dyes have different densities.  Also, different colors may transmit more or less light than the other two.  For this reason, the only practical way to mix the starch is by trial and error, and heavy emphasis on taking notes.  

Firstly, let's coat a glass plate with the first varnish.  This will give us a substrate on which to dust the starch we're testing.  With the glass plate in one hand and the bottle in the other, pour an amount of the varnish onto the center of the glass.  Gently rock the plate around, so that the varnish covers the whole surface.  When done, tip the corner of the plate and let the excess run back into the bottle.  Set the plate down on a level surface and allow an hour or so for it to be completely dry -- if you can still smell the xylene, it's not dry yet!  

While you're waiting on this, measure out 2g of each color of starch into a beaker (or whatever other container), and mix it around with a small spoon or spatula.  Eventually the colors will disappear, and the mix should take on a dark red appearance.  You can't overmix, so be thorough here!  

When the plate has dried, scoop a dime-sized amount of starch onto the plate, and spread it around a bit with a soft paint brush to about the size of a quarter.  Brush the extra back into your starch container.  There will be a lot of interstitial space between the grains, so to get a proper idea of the color balance we're going to need to crush it.  This can be done a few ways:

1. Take a spoon and just kind of roll it around for a few seconds.  You should notice the smushed starch visually darken a bit, and become a bit more shiny.

2.  My preferred method is to take a 1" castor transfer ball (like these lil duders here) and roll it around in a few small circles while applying a downward pressure.  You might want to try and wipe some of the oil off of the ball first, as it'll soak into the starch a bit (it's not a huge deal though).  


After a small section has been crushed, observe it against a nice cool light, phone screen, or diffuse sun on a white object.  You should notice a color cast.  From here, you will have to determine what color to add and start the process over.  Try adding 0.5g of whatever corrective color needs to be added, remix, and flatten again.  This can take several tries until the screen becomes close to neutral.  Don't get impatient, but don't panic if it's not exactly neutral either.  Each batch of screens I've made always have a slight color bias, and this can be corrected for later.  Keep track of the additions, and once you're happy with your color balance, scale it up to make about 50-100g of starch mix.

Now that we officially have our starch ready, we can begin to work on our first autochrome screen plates! On several 4x5 (or whatever is your preferred size), take 1/4" black masking tape and mask the edges of each of the plates.  From here, we repeat the same procedure we did earlier when testing the starch mix.  First, pour on the first varnish we mixed up, rocking the plate for full coverage, and then pour the excess back into the bottle.  Allow the plates to dry completely for a few hours on a level surface, until they no longer smell like toluene. 

After they have dried, take a quarter-sized scoop of starch and plop it in the middle of the plate.  Spread the starch around with a small paint brush until it has covered the whole plate.  With a soft, long-haired makeup brush, gently brush the extra off the plate and back into the starch container.  Now we should be ready for compression.

Part 5 - Compressing the Starch


The starch is now adhered to the sticky plate surface, but there is still quite a bit of interstitial space in between each grain.  Not only will compression reduce this in-between space, it will also considerably improve the plate's overall transparency. 

The compression step can be frustrating, time-intensive and somewhat difficult, but it's definitely far from impossible.  While I use a modified CNC router to do most of the hard work these days, good results can be achieved by hand, and fairly inexpensively to boot.  I'll discuss the ins and outs of each, starting with hand pressing.

Hand Pressing

I found out early on that relatively satisfactory results could be easily achieved just by smashing the starch with a spoon.  However, this way proved to take a ridiculously long time to cover, say, a 4x5 plate.  I tried a few things, including knitting needles and those 10mL rollerball bottles used for perfume and essential oils.  My best results were achieved with the 1" transfer ball that I mentioned earlier in Part 4.

I like to wipe the excess oil off the castor ball with a paper towel first before getting started.  Gripping the castor ball in your hand, start rolling in on the plate in small little circles, slowly working your way across the whole plate.  This whole step is very reminiscent of shading in a drawing with a pencil.  In my experience, small circles help avoid the ball bearing seizing up.  I've noticed that longer lines will often cause the ball to lock up, and usually ends up in the bearing scratching a large streak.

You can expect this to take about 30-45 minutes for a 4x5 sized plate.  It's a painfully slow process when you have a lot of plates to do -- I used to do a little bit of pressing whenever I was waiting on something to cook, or needed a break from work.  Doing a bunch all at once is apt to make your fingers sore!

The following plates were all made with hand-pressed starch:

You can expect this to take about 30-45 minutes for a 4x5 sized plate.  It's a painfully slow process when you have a lot of plates to do -- I used to do a little bit of pressing whenever I was waiting on something to cook, or needed a break from work.  Doing a bunch all at once is apt to make your fingers sore!

Machine Pressing 


Pressing the starch by machine is infinitely less tedious and cleaner results can be obtained.  With a roller and CNC setup like I've been using for the last year or so, a 4x5 plate can be pressed in about 10 - 15 minutes, and is much more of a "set and forget" type of process.  


Initially I used a Shapeoko 3 CNC router to press plates.  The router is fully belt-driven, which means it's motion is very quick compared to an axis with ballscrew.  I'm not sure if this was a problem limited to my particular model, or if all Shapeokos suffer from this, but unfortunately the z-axis belt would skip if too much downward force was applied.  This limited roller to 1) have an extremely small contact area and 2) require multiple passes to achieve a satisfactory pressing.  In general I found that the plate would require about 14 passes to achieve a satisfactory degree of compression.  With this style of roller, it took about 45 minutes to press a 4x5 plate.  

The roller was made from a window guide-roller, with a 1/4" bolt as an axis.  The sides of the roller were sanded down to be a little less sharp, as it tended to cut into the plate and leave big lines.

Next, I tried a lower-end 3040 CNC machine that uses all ballscrew axes.  The overall travel motion is slower, but the ballscrew z-axis allows the machine to apply more downward force on the plate.  I've iterated through a few different roller designs, but have had the best consistent success with a polycarbonate one designed for a Creality Cr-10.  The frame is made from laser cut 1/4" plywood, and uses an M5 bolt as an axis.

To avoid sharper edges of the wheel cutting into the plate, the whole wheel will need to be sanded carefully.  

As mentioned, despite the fact that the machine travel speed is overall slower than the Shapeoko 3, the machine can apply more force to the plate, allowing the use of a wider roller and a single-pass motion across the plate.  Typically a 4x5 plate can be fully pressed in about 15 minutes.  Care must  be taken, since this machine is able to apply enough force to crack the glass!  An unattended machine will quickly pulverize the glass, damaging the roller. 

I home the x and y axes by jogging them up/left until no more motion is permitted and the stepper motor skips steps.  Similarly, I home my z axis by jogging down into the plate until the motor skips.  The axis cannot crack the glass without rolling back and fourth across the plate, so no worries here yet.  The overall pressure applied by the roller can be varied by changing the z height in the program.  For my machine, z=0.8mm is about as much pressure as a 4x5 plate can take before cracking becomes common.

The base the plate rests on must be fairly rigid, as any amount of flex will crack the glass.  MDF worked well for a while for 4x5 sized plates, but 5x7 sized plates would routinely bend too much and break apart.  I changed the substrate to a thick piece of glass (1/4") which was decidedly less flexible, with improved results. 

If anyone wanted to try out machine pressing, I'd definitely recommend going with the 3040 version, rather than the Shapeoko.  It's much cheaper (~600USD usually) and can reliably press plates much more quickly. 

After completing the pressing, I always laser engrave the edges of each plate.  Numbering them has helped me a ton while experimenting with different second varnish coatings. 

Part 6 - Applying the Second Varnish

The "Second Varnish" serves as a final protective layer to seal in the potato starch, protecting the grains from processing chemistry.  Even trace amounts of moisture that may find its way into the starch layer will cause the colors to fade and run, so a proper seal is important!  Traditionally, the Lumières used a nitrocellulose based varnish, dissolved in ethyl acetate, with a secondary component of damar resin that gave the coating a grip for the gelatin to stick to.   I've tried (with mostly failure) trying to adequately replicate this varnish.  Though I'm not 100% certain yet, it seems like Golden's MSA Varnish seems like an adequate modern replacement. 

This section may be apt to change in the future, as the Golden varnish appears susceptible to a problem with "desaturation", which does not start to set in until months after the autochrome has been completed.  This will be discussed more in depth toward the end of this article.

To prepare the second varnish, simply dilute the Golden MSA varnish 1:1 with artist's grade turpentine.  I assume this dilution is stable, but I've always prepared it as needed, and discarded the leftovers.  Dilution achieves two things:  Firstly, undiluted, the Golden varnish is incredibly viscous and difficult to work with.  Second, the dilution allows for an incredibly thin and even coating to be applied after the solvent has evaporated.