Working with multiple masters

I've never worked with multiple masters.
I understand the theory (I have background in math which helps...), but I feel like a tutorial that focuses on the workflow, practical advice, etc., could help.
Do you know if and where I can find something like that?
(Tutorials for specific font editors are also welcome.)


  • Paul HanslowPaul Hanslow Posts: 11
    edited August 2019
    This subject can become a bit of a minefield if you venture into the Variable Font territory and how multiple masters function (especially when dealing with multiple axis such as weight, width, optical side, etc). Some of this not even I fully understand.

    However, there are a few resources which can help understand multiple masters. Some better (and some drier) than others. As it's uncertain what skill level you're at, here are some general links which I hope you'll find helpful:

    1. Tutorial on setting up Multiple masters (Glyphs app)

    2. Designing Multiple Masters typefaces by Adobe (General Theory)
    It's a longer read and I've never read it as I learn from doing more so than reading. 

    3. Multiple Masters (FontLab5 app) 
    a little old as FL6 is out, but good if you're a visual learner.

    In regards to workflow, this will be dependant on the designer, the job, and what application(s) they feel most comfortable using. For example, I prefer setting up masters in Glyphs than FontLab 6, unless working on a variable font as FL6 provides better visual feedback in this regard. Other type designers may prefer the opposite to myself. 

    It may help to provide information on what area(s) of multiple masters intrigue you the most. Is it the mathematical theory side or a practical/applicational interest?
  • James PuckettJames Puckett Posts: 1,665
    2. Designing Multiple Masters typefaces by Adobe (General Theory)
    It's a longer read and I've never read it as I learn from doing more so than reading. 
    Definitely read this book. And the relevant tutorials for whatever program you’re working with.
  • The mathematical theory sounds very simple. It's just convex (or affine, if you allow extrapolation) combinations of points in the plane, isn't it?

    I'm interested in the practical side. I thought maybe I could add an axis or two to Arrowwood, mainly just to play with it (I'm not sure where this project is going anyhow, I hadn't planned to take it beyond that logo).

    Thanks for those resources!
  • Glyphs uses the terms Multiple Master but is not strictly following the Adobe MM structure. You can place master much more freely and don't need to have a master at each corner of the design space. So you might have a Light, Bold and Regular-Condensed. The Glyphs handbook has a chapter about his:
  • Thanks, Georg. I've actually already looked in the handbook. I'm not looking for answers to specific questions. I was sort of hoping to find something about how to best approach my first experience with multiple masters on the most practical level: workflow, tips, and so on. Maybe there's not much to say about that and there's no reason why something like that should exist in the first place...
  • @Ori Ben-Dor: There is a VAST amount to say on it!

    “Multiple Master” was a specific Adobe format, used with Type 1.

    MM had a limit of 16 masters for a font. And the most important thing is that although it was never a theoretical limitation (!), as actually implemented, MM required a master in every corner of the design space. Which effectively gave MM fonts a limit of 4 axes.

    MM in a later revision allowed for intermediate masters, but the MM version of this was never supported in most tools.

    OpenType Variations (variable fonts), based on Apple’s old GX Variations technology, is much more flexible. There is a limit of 64K axes. I do not recall the limit on the number of masters, but it is similarly huge.

    Importantly, variable fonts place few limitations on the placement of masters in the design space. However, due to the math of vector addition and interpolation, there are some real complexities about where masters can be placed and have them both: (a) take part in interpolation, and (b) yield at all predictable results.

    The easy part is to say that:
    1) There is a default master. Each further master must be connected, via a single orthogonal move, to another master, and every such chain must connect to the default master.
    2) Having masters in every corner of the design space is a simple, legal approach. Like the old MM. If you do this, everything is done by interpolation. But the number of masters is the square of the number of axes. Imagine a cube in which each corner is a master; that is an example of this approach.
    3) The main alternative is to use vector addition. In this case, you have a "star" in which each axis is defined by a master sprouting from the default master. Imagine a cube in which the default is at a corner, and the three adjacent corners in each direction each have a master. So four of the eight corners of the cube have masters (but not all in one plane!). The number of masters is 1+n, where n is the number of axes. So for three axes, instead of eight masters with the MM approach, you only need four. BUT your control of the design space is a lot less direct.

    You can also do the star with axes sprouting in each direction. That is, the default master is in the middle, and some or all the axes are defined by a pair of masters that sprout in two opposite directions.

    Similarly, you can glue a pair of cubes together, sharing one side, and have even more control by using twelve masters instead of eight.

    I am describing 3-axis variations, but all this applies to more or fewer axes as well. But three is moderately complex, and is the most our human brains can easily visualize, so it is a convenient number.

  • Thomas, thank you so much for this comprehensive review, it's very insightful!
  • Thomas PhinneyThomas Phinney Posts: 1,726
    edited August 2019
    You can do hybrid approaches. For example, you can have a cube of eight masters, with one corner sprouting a star. Or to put it another way, a star-based 3-d configuration in which one of the four quadrants of the design space is defined by a cube with no extrapolation, and the other three quadrants use vector addition.

    I have an idea, to do a video using a TinkerToy construction set to show and explain this stuff. At least in the 2D-3D versions. I think I have a sponsor for doing a pretty nice version of this. (But it will have to wait some months. Must finish Science Gothic.)
  • I think I got it, but I can see how a video could help explain this stuff :smile:
  • I am getting better at explaining it in writing, after a lot of practice.  ;)  But I think a video would be really good. Or at least an article with visuals, but 3D works better with, well, real 3D stuff.

    Then again, I could just build 3D graphics that work with WebGL and put the whole thing in an article. Hmmmm.
  • There is a segment about visualizing interpolation spaces in Luc(as) de Groot's talk at TypoLabs 2018 and you can also try to drop a font on the Multidimensional Axis Visualizer shown in the talk.
  • What is especially tricky is not just the visualization, but understanding which master/axis configurations will work the way one expects, and which will “fail.” I did not find this at all intuitive, at first. 
Sign In or Register to comment.