Welcome to Librarium Online!
Join our community of 80,000+ members and take part in the number one resource for Warhammer and Warhammer 40K discussion!
Registering gives you full access to take part in discussions, upload pictures, contact other members and search everything!
You may have seen this on some of the other forums and websites over the past few days - but, for anyone who hasn't...
This is a little program that I have been working on for awhile now that cross references a few dozen different paint lines to figure out which ones are suitable replacements between the lines.
Color Match 1.0
This first tool allows you to compare a single paint or color to multiple paint lines. You can filter the results to show flats, glosses, metallics and others as well as find cheaper alternatives in craft paints or aerosols for doing larger areas.
Full Line Comparison
This option lets you compare two complete lines to one another. It only matches similar paint finishes to one another though (gloss to gloss or metallic to metallic for example).
This will give you a bit of background information regarding the tools and process behind how the samples are obtained as well as the formulas used in order to make the comparisons.
How it is being used.
Short article that I wrote to explain some of the different things I have found the tool to be useful for.
Any questions, comments, suggestions?
Thanks for the link. A very useful guide. I haven't read all the technical documentation you posted yet but it seems quite interesting.
Generally speaking, how accurate is it? Just curious, as I received a few inconsistencies when testing it. I also found that the results didn't match the widely accepted conversion charts that have been posted around the web...
Once again, a useful chart, but I've got to try get through your documentation before doing further testing and gleaning a thorough understanding of the application.
(yes... I've been involved in IT for many years... dev, testing, admin, analysis, etc...)
A very useful link... Thank you... I haven't noticed anything wrong but I was just trying Vallejo to GW and vice versa...
Thanks to all the people showing faith in me
I sometimes feel that everyone will leave me and they will be on their own...
the interface on the first link is slightly confusing. it took me a while to get a paint comparison going. My recommendation would be to put a set of numbered steps (do this first, then do this) and flag whatever is optional as such.
you could write a "closeness" algorithm to figure out what the best matches are across multiple lines, and recommend the top 3 for example.
Also, a nice touch would be to get the pantone values as well, since that's what's most popular in the paint/print world.
another idea would be to allow for paint vendor groupings, either by medium (acrylic vs enamel) or by manufacturer (get all the vallejos together if you know a store that carries them)
looking at the comparison between Vallejo Game Color and GW Citadel paints (not foundation) the matchups don't match what vallejo posts on their website here:
here's the chart where they compare Vallejo Game Colour to Reaper, GW and Rackham
I think one limitation of the process you took (as described in your FAQ) is the colour of your primer/surface when you are making your paint samples.
Also, how are you calibrating the scanner before doing the first scan? Have you considered using 3 or 4 scanners to remove systemic bias? I know for example that my flatbed scanner tends towards blue when I'm scanning, yet my digital camera over saturates reds.
finally, you have no way of correcting for differences between batches of paint. I've seen quite a bit of variance between two bottles of Vallejo bleached bone for example. this is not something that's really possible to correct for without some sort of community based approach, and then you can't control the paint prep process.
Definitely a very ambitious project. I'm constantly surprised by the level of secrecy that surrounds colours, from teh pantone trademark issues, to the fact that I still can't get a decent match for my Ralph Lauren wall colour in a cheaper paint.
Here's another chart that compares GW to Vallejo Model and Game color lines:
I've found this one to be ok for accuracy when actually applying to the models, but there's definitely very few *identical* colours in teh Model Color and Game Color lines.
However the samples are not necessarily uniform. Due to the nature of paints (solids in a liquid suspension) bottles of paint will vary from batch to batch as well as how full the fill container was when the bottle was filled (paints from the bottom of the barrel tend to be more pigment rich than those from the top). There are also issues relating to organic pigments versus oxide pigments. Oxides tend to remain colorfast for years - even decades, while the cheaper and more commonly used organic pigments generally will fade with time due to oxidization and exposure to UV. This becomes an issue because some of my paint samples are from the first color chip sheets I did nearly 10 years ago (I am working on fixing that though).
You then get into additional issues with the actual samples. If the paints are not fully mixed, they may not contain a uniform dispersion of pigments. This will lead to issues of course. Since I don't own each and every paint in the database (over 3000 today with another 100 or so coming this weekend) I have limited control over the samples that are sent to me by people who want to help out. I do explain what needs to be done, but I can't verify that it has been done. Related to the samples is the issue of transparent paints and inks. Although I have made every attempt to get samples from color dense areas of the samples it is sometimes very difficult. One obvious example of this is GW Citadel Red Ink. You can see that the color is washed out somewhat - this is a result of the white substrate showing through the sample. Again, issues that I have been working on a better sampling method in order to avoid the problems related to them.
The final issue are web colors. There are around a total of 200 paints that I currently do not own or have a physical paint chip sample. After a fair amount of thought, I have chosen to use the samples provided by the manufacturers as the starting point for those samplings while I arrange for a physical sample (either by tracking down someone who owns the paint or ordering it myself). Now I don't just take the sample from their website and use it straight away. I compare their online samples to physical samples in order to determine what color shifts might have occurred while they determined their data. More often than not, an obvious shift exists, and I correct it to be inline with my samples. While not a perfect solution, I find it better than leaving a hole in the data sets. The paint line probably most affected by this issue are the Gamecraft Acrylic paints. Currently there is no US supplier for them, and getting physical samples has proven to be quite difficult.
In regards to the colors being in disagreement with the accepted online lists...I wouldn't doubt it at all. In fact, it is to be expected. I cover the issues related to human color perception a bit in the FAQ - but let me show you an example of why I don't worry much at all about what the other lists are showing.According to the official list provided by Vallejo, their match for Citadel Warlock Purple is Warlord Purple. In the same listing they say that the Reaper Master Series match for their Warlord Purple is Pale Violet Red. Now if you take a look at the official matches put out by Reaper, you will see that their official match for GW's Warlock Purple is Clear Magenta. Since the professionals can't seem to come to an agreement regarding what paints are supposed to be the same, I figure I'll go ahead and give it a shot using a bit of science and programming as opposed to subjectively looking at the colors myself.There are a lot of reasons why those lists may not jive well together, but all of those issues exist in pretty much every list that you will find...some are even more pronounced than others.
That took a bit more space to write than I planned...
Once you figured out the magic combination though, everything flowed OK though, right?
The Full Line tool will run into a wall though. Once I get the "Official Match" option implemented, it will be trivial to provide all the official matches (based on manufacturers recommendations and what not), however due to the math involved I need to be careful about just letting loose the script to compare all the colors in the database. While it isn't a huge issue for the smaller paint lines, the test script that I have on my development server starts to bog down when I run through a big line like Reaper Master Series or Vallejo Model Color. When you add into the mix problems with internet delay and multiple users hitting the database at the same time, I have been getting a few timeouts from the server.
It is an issue that I am still working on, and I do agree that it would be great if I can get it workable - for now though, there are a couple things which are slightly more pressing issues.
Eventually I would like to be able to add those (Pantone and other color standards) as well, but for now that is a back burner project that I have been casually pursuing.
I think I addressed your concerns regarding the accuracy fairly well above.
An additional issue which will come up in the future is the issue of scale color. Once the FS fandeck gets added to the mix, people who are doing historical models and miniatures will find that Tamiya and Testors preadjust for scale color. This is the property of color which causes it to look darker when viewed from further away (modelers and historical miniature painters will often use scale color to help reinforce the idea that you are looking at a real model...just from really far away). Anywho, the result though is that when you compare a paint sample that is supposed to be a specific FS color, it looks darker than the FS sample. I haven't had a chance to check to see if Vallejo does the same with their Model Color line though.
In this case I am not too concerned with an absolute color accuracy as I am an absolute color uniformity. If my scanner is off on one sample, it will be off on all of them. While this may affect the way the color is displayed, it shouldn't affect the underlying mathematical comparisons.
Right now I have been receiving paint samples from around 2 dozen other hobbyists (hopefully more now that it has gone fully public). Quite often, I get three or four samples for the same paint. When this happens I scan each separately and than the results are loaded into the database as unique datapoints. The program than averages those datapoints and uses that average as its point of reference for the comparison.
Although I do not have full control over the final sample, I do go as far as I can to keep things uniform. For people who want to contribute, I prep a sample substrate and send it out to them with instructions for applying the sample. If they would rather send a wet sample (becomes a bit of an issue because it is very hard to properly mix two or three drops of paint), I arrange for that as well. Not perfect...but working to get pretty close.
Also, when a color is brought to my attention - I go back and double check things, a few have been wrong due to typos on my part...it happens. If it is a single sample, I'll go ahead and purchase three bottles of paint from different suppliers in order to increase my chances of getting different batches. I'll then go ahead and sample those in order to get a better representation of the paint.
Hope that helps to address some of the issues you brought up. Let me know if you have any other suggestions or ideas.
Last edited by Joe Kutz; November 16th, 2007 at 18:56. Reason: Added thoughts.
Thanks for such a comprehensive reply. Looks like you may have had to field these questions before
Hope you don't mind, but I posted a link to your site and tool on my local haunt, here's the link to my post: post on HF
In a week or so (still have one large announcement that will likely cause a lot of questions), I'm going to go through and compile all the questions again for the FAQ...so that should help too.
Hey, this is really helpful, thanks for putting it up. Rep for you!