[OTVar] Contrast Axis (ctst)

Belleve Invis

Hi all,

I am writing the proposal about contrast axis for @Peter Constable. However the biggest problem is the way we interpret the value.

My general idea is:

1. Higher contrast should reflect higher ctst value;
2. Positive value for "normal" contrast;
3. Zero for contrast-less sans-serifs;
4. Negative for decorative "reverse-contrast" fonts;

But the quantity become a problem. There are two possible ways:

1. Absolute manner: use some formula to transform the hairline-stem ratio into ctst value;
2. Relative manner: value 100, 0 or -100 is always the "normal" contrast, and the ctst value is a relative quantity.

I have no idea which way is better. So do TypeDrawers folks have any idea about it?

Thanks,
Belleve

Adam Twardoch

Using an actual measured ratio between the thins and thicks is not a good idea: if a font has a weight axis as well, then "high-contrast light" will have a lower thin-to-thick ratio than "high-contrast black", simply because the denominators are much larger in blacks. 

Also, the scale should quite possibly express perception rather than geometry (even ITC Avant Garde doesn't have the thin-to-thick ratio = 1). 

I like your proposal for a somewhat arbitrary scale: 

- 0 should mean: the glyphs have no perceived contrast
- increasing positive numbers should mean: the contrast increases towards what is commonly known as "traditional" contrast for a given writing system
- decreasing negative numbers should mean: the contrast increases towards what is commonly known as "reverse" contrast for a given writing system

-100 and 100 should define what is broadly seen as "extreme"
contrast and are the minimum/maximum permissible values. 

Typical reading-size contrast should be <33 (typically around 20-25). 

Not sure how the scale could be substantiated further. 

LeMo aka PatternMan aka Frank E Blokland

Hi Belleve,

Just curious: if the weight axis covers the increase/decrease of the bold parts while (more or less) keeping the thin parts identical and the optical-size axis covers the increase/decrease of the thin parts while (more or less) keeping the bold parts identical, how exactly does the contrast axis relate to, and interact with these axes?

Best, Frank

Hrant Հրանդ Փափազեան Papazian

I myself would simply offer the following terminological advice:
— Whatever you do please don't use the term "stroke" anywhere, because that can limit cultural exploration hence progress.
— Avoid "reverse contrast", for two reasons: it's Latin-centric, and casts a negative light on many non-Latin scripts; and it saddles horizontal contrast (the term I encourage) with an undesirable stigma (as do "decorative" and "negative" BTW).

Mark Simonson

I don't see the problem with the term "reverse contrast". It could be taken as the reverse of normal contrast, whatever that may be for a particular script. It doesn't have to mean thin verticals and thick horizontals just because that's what it means for latin fonts.

Hrant Հրանդ Փափազեան Papazian

That seems like an elegant stance, but there's still the potential of people seeing scripts with horizontal contrast as abnormal. And what about scripts that conventionally have no contrast, like Thai? You can't call a Thai font with thicker horizontals "reverse-contrast" without referring to Latin, and furthermore making it sound somehow inferior. Lastly, it can make things pretty confusing in a multi-script system with mixed contrast...

In the end that "normal" is the problem; I see no reason to give vertical contrast in Latin such an unassailable position. Calling something "reverse" implies it's inferior, and that impedes exploration. In contrast :-) "vertical" and "horizontal" have no baggage.

LeMo aka PatternMan aka Frank E Blokland

The question is: for whom is the proposed contrast axis meant? The contrast in a typeface is not arbitrary, i.e., it is part of the design, and if there is a tolerance then this will be mostly related to the point size. The optical-size axis makes clear to the enduser that the contrast range is not arbitrary but related to size. However, the contrast axis seems to suggest that within any type design this range can be huge (perhaps I misinterpret the proposal). Also, it will put the control over the range unrestrictedly and without any indication other then number ranges in hands of the enduser, who can mess up things completely by, for example, not hampered by too much knowledge selecting a very high contrast for small point sizes in print.

I think the biggest danger for the variable-font model is that font developers make it a tool for their own technical satisfaction. If this results in proprietary axes, the problem is limited to the fonts in question, but I think we should be very careful with registered axes. When it comes to proprietary axes it is not difficult to open Pandora’s box. I am thinking, for example, of ink squashes (a contrast-lowering axis) and the rounding of corners for revivals. Ink traps could be another axis.

Jens Kutilek

I agree with Frank that the amount of contrast is an integral part of a typeface design.

If a value of 0 (no visible contrast) is the default, as Adam suggests, will we have to design monolinear versions of all typefaces? Will it be reasonable to assume that any typeface should be able to change its contrast completely from conventional to reverse? When is a typeface not itself anymore?

Regarding the name "Contrast Axis": There may be other axes related to contrast, e.g. contrast direction (angle of the thick and thin parts) or contrast type (translation, expansion). This should be considered when naming an axis. Would "Contrast amount" be a better name for the proposed axis?

Belleve Invis

@LeMo aka PatternMan aka Frank E Blokland

I mainly work for CJK and there are really some fonts with adjustable contrast. TP Mincho is an example:

Masaya Nakamura

@Belleve Invis Thank you for referring TP Mincho. You can try it on our website, in a pre-OTVar-age manner.

As I wrote in another thread before, some Japanese families have contrast variations. Notably, a famous typeface Typos, which has been released since 1962, has a numeric naming scheme. It's based on the horizontal/vertical stem width by the 1/100 em:
（桑山弥三郎「『タイポス』『タイポスオールマイティ』と最近の新書体」p. 77、『タイポグラフィ・タイプフェイスの現在』女子美術大学、2007）

Hrant Հրանդ Փափազեան Papazian

Everything is an integral part of a typeface design.

Adam Twardoch

Jens Kutilek said:

I agree with Frank that the amount of contrast is an integral part of a typeface design.

If a value of 0 (no visible contrast) is the default, as Adam suggests, will we have to design monolinear versions of all typefaces? Will it be reasonable to assume that any typeface should be able to change its contrast completely from conventional to reverse? When is a typeface not itself anymore?

I don't think I said 0 should be "default". Have I said this? I thought I said 0 should be "perceived no contrast". 

I also said that it might be useful if the scale is calibrated so that the generally perceived (decent natural contrast for text use) — if there is contrast — should be somewhere in the range 25-33. 

So that different designs are coarsely comparable. But of course if you  design a variable font where there is always contrast, just slightly more or less, then if it has a ctst axis, it could vary between, say, 18 and 47. Another could vary between -5 and 5, and another between 80 and 92. 

I also said that positive values should be associated with contrast that is the traditional contrast for a given writing system. Most writing systems, if they have contrast, have some sort of "natural" or "traditional" contrast direction. It doesn't matter if it's not  vertical or more horizontal, it can also be up to the designer's interpretation. 

All writing systems have "reverse contrast" — with "reverse" meaning "opposite axis to what you would traditionally expect". Hebrew has a traditionally more vertical contrast, and Latin has traditionally a more horizontal contrast.

And if you, as a designer, think that your contrast is "naturally correct" and don't wish to treat it as "reverse", then you can still use the positive numbers. So for Greek or Arabic, designers would have some leeway — some could decide that their "vertical-contrast" Greek is "natural" or "traditional" and use positive numbers there but perhaps also go into horizontal contrast using negative numbers — but others could just increase contrast horizontally with positive numbers and not care about the negative bales at all. 

No witches and no stakes here. I'd treat the negative value space largely as "reverse — whatever you as a designer think it is and as long as you choose to use that concept". In my view, the vast majority will only use the positive space. 

As far as trying to come up with a slightly universalized scale — this is also a convention, and a tool for the designers who don't have a clear opinion that they want to do it differently. Everyone is always free to break the rules consciously — and then accept the consequences. 

Adam Twardoch

(As with any axis, the default should be what the designer feels should be the default. If the slnt axis goes between 4 and 12 in a design, the default can be 4 or 12 or 11.30993 (the last being a rational angle of course)

Adam Twardoch

Before I forget: here’s a list of “beauty angles” or “rational angles”, i.e. the hypoteneuses of a right-angled triangle where both legs are small integer numbers. With those angles, node and handle coordinates will have a much higer chance for proper integer rounding, while really keeping the actual slope. I should post it somewhere else, because this is off-topic, but I will forget so I’m doing it here.  

| Italic angle | Caret run | Caret rise |  
| ---          | ---       | ---        |  
|  0.00000     | 0         |  1         |  
|  2.86241     | 1         | 20         |  
|  3.81407     | 1         | 15         |  
|  4.76364     | 1         | 12         |  
|  5.71059     | 1         | 10         |  
|  6.34019     | 1         |  9         |  
|  7.12502     | 1         |  8         |  
|  8.13010     | 1         |  7         |  
|  9.46232     | 1         |  6         |  
| 11.30993     | 1         |  5         |  
| 12.52881     | 2         |  9         |  
| 14.03624     | 1         |  4         |  
| 15.94540     | 2         |  7         |  
| 16.69924     | 3         | 10         | 

Jens Kutilek

Adam Twardoch said:

Jens Kutilek said:

I agree with Frank that the amount of contrast is an integral part of a typeface design.

If a value of 0 (no visible contrast) is the default, as Adam suggests, will we have to design monolinear versions of all typefaces? Will it be reasonable to assume that any typeface should be able to change its contrast completely from conventional to reverse? When is a typeface not itself anymore?

I don't think I said 0 should be "default". Have I said this? I thought I said 0 should be "perceived no contrast". 

Yeah, I guess I did read your numbers as internal axis coordinates, with 0 being the default.

Though a typeface may span only a small range of the design space, I wonder if new axes will raise expectations as to what a typeface can do?

I have the feeling that for width and weight axes, we as type users already expect everything to be simply available, from Hairline to Black and from Ultra Compressed to Ultra Extended. So there may be a pressure for type designers to span "the whole axis range" in a new design.

Jens Kutilek

Hrant H. Papazian said:

Everything is an integral part of a typeface design.

Of course. But I think there are differences in perception of all the variable parameters available today. I can only speak for myself, anecdotal evidence, that I still see extreme weight and width changes as valid variants of the typeface design (say e.g. of the Regular style), but if only the contrast changes by a quite small amount, it quickly starts to look like a different typeface design to me.

That is not to say that a contrast axis could not be useful.

Adam Twardoch

Jens Kutilek said:

I have the feeling that for width and weight axes, we as type users already expect everything to be simply available, from Hairline to Black and from Ultra Compressed to Ultra Extended. 

Do we? I never have. On the contrary, from the very beginning I said that foundries should consider small-range variable fonts, and that some may do better if they upgrade their families in a backwards-compatible way: if you offered 6 weights, you can still offer 6 variable fonts, each centering on the old static weight but providing a weight axis that covers a section of the full axis so that only all 9 would cover the entire space. Or maybe offer 3 such fonts, and if you have a big family such as FF Good that currently ships as 49 upright and 49 Italic fonts, you could replace it with 9+9 fonts. 

Adam Twardoch

Ps. I don't think users expect that each font covers all of Unicode.

Adam Twardoch

Even David Berlow hasn't yet combined Decovar and Amstelvar into one font (but I'm sure it'll happen soon!).

Thomas Phinney

Jens Kutilek said:

I agree with Frank that the amount of contrast is an integral part of a typeface design.

If a value of 0 (no visible contrast) is the default, as Adam suggests, will we have to design monolinear versions of all typefaces? Will it be reasonable to assume that any typeface should be able to change its contrast completely from conventional to reverse? When is a typeface not itself anymore?

Regarding the name "Contrast Axis": There may be other axes related to contrast, e.g. contrast direction (angle of the thick and thin parts) or contrast type (translation, expansion). This should be considered when naming an axis. Would "Contrast amount" be a better name for the proposed axis?

Sure, it's an integral part of typeface design. So are many other things that we give axes for. Not all axes need to be in all fonts, either.

And I agree with your point that there may be other contrast-related axes. We should either know what those would be named, or at least take their possibility into account, when naming this one.

“If a value of 0 (no visible contrast) is the default, as Adam suggests, will we have to design monolinear versions of all typefaces? ”

- That wasn't Adam, it was Belleve (unless Adam edited his post?)
- If all typefaces had a default of zero, then zero would mean "default" and not "no contrast" (which is a good argument against both those concepts, IMO). You are mixing two incompatible options that nobody else has proposed mixing, IMO.
- Nobody has suggested for this axis that any particular functional range is required, that I have seen.

Belleve Invis

Thomas Phinney said:

Jens Kutilek said:

I agree with Frank that the amount of contrast is an integral part of a typeface design.

If a value of 0 (no visible contrast) is the default, as Adam suggests, will we have to design monolinear versions of all typefaces? Will it be reasonable to assume that any typeface should be able to change its contrast completely from conventional to reverse? When is a typeface not itself anymore?

Regarding the name "Contrast Axis": There may be other axes related to contrast, e.g. contrast direction (angle of the thick and thin parts) or contrast type (translation, expansion). This should be considered when naming an axis. Would "Contrast amount" be a better name for the proposed axis?

Sure, it's an integral part of typeface design. So are many other things that we give axes for. Not all axes need to be in all fonts, either.

And I agree with your point that there may be other contrast-related axes. We should either know what those would be named, or at least take their possibility into account, when naming this one.

“If a value of 0 (no visible contrast) is the default, as Adam suggests, will we have to design monolinear versions of all typefaces? ”

- That wasn't Adam, it was Belleve (unless Adam edited his post?)
- If all typefaces had a default of zero, then zero would mean "default" and not "no contrast" (which is a good argument against both those concepts, IMO). You are mixing two incompatible options that nobody else has proposed mixing, IMO.
- Nobody has suggested for this axis that any particular functional range is required, that I have seen.

I do not said that 0 should be default. In my draft the default value could be either -100, 0 or +100, depending on the design of "normal" style.

Thomas Phinney

That would be with your second method. But in full, that means that there is no consistent meaning of the particular numbers, across different fonts. That seems like a bad thing, to me. So I would prefer the first approach.