Random Point Movement facility of the Glyph Transformer

[William]

I noticed that the Glyph Transformer has the Random Point Movement facility.

I have been having some fun trying out some tests of what effect is produced.

As there are no parameters to set or choose, I wonder if you could possibly say please what is the algorithm for doing the movement?

For example, is it simply that each point can move up to one hundred font units horizontally and vertically with an ordinary random number generator, with no regard as to how far away other points are located, or something else please?

I tried a contour with a vertical line which started off with on-curve points (1024, 200), (1024, 300) .... (1024, 1800) then points at (1536, 1800) and (1536, 200) to close the contour. I tried the Random Point Movement time after time, noting that sometimes a result, after several transforms, had overlapping contours.

It is fun experimenting yet I am wondering what is the intended use of the Random Point Movement facility. Any ideas please?

William Overington

2 March 2005

[Erwin Denissen]

A random number between -50 and 50 will be added. As you already noticed it is fun, that's all.

[William]

Thank you.

William

[William]

Although Random Point Movement has no parameter choice available, I have found that one can effectively have a smaller range of random number by using the scale transform to enlarge the glyph with respect to the point (0,0), then apply the Random Point Movement transform, then use scale again with respect to the point(0,0) so as to get back to the original size.

For example, scale by 200%, Random Point Movement, scale by 50% so as in effect to use random numbers in the range -25 to +25.

I have been experimenting with using the basic Random Point Movement transform.

I produced a basic lowercase h using only horizontal and vertical lines. The vertical construction lines were at 136 (=256 - 120) font units horizontally, 376 (=256 + 120) font units, 648 (=768 - 120) font units and 888 (=768 + 120) font units. The horizontal construction lines were at 1024 font units vertically and 784 (=1024 - 240) font units. The h is 1792 font units high, with its base at 0 font units high. I made the corners of the letter to be on curve points and then I added off curve points every 128 font units, as long as an off curve point was at least 128 font units from any other point. Thus there would be no overlapping contours after using the Random Point Movement transform.

I then copied the font to a new file, using Save As, so as to keep the original intact.

I then applied the Random Point Movement transform. The effect is interesting. It is as if the letter is made up of ripples in water.

An interesting point though is the matter of adjusting the bearings. This starts at 136 font units. After the application of the Random Point Movement transform that changes. I have found that the best result is obtained by manually looking at the properties, finding the leftmost figure of the Glyph Boundings and manually setting Left Side Bearing to that value before saving the font to disc.

With a font with many letters this may mean quite a lot of effort in manually setting the Left Side Bearing for each character. At present I am unaware of that process being possible to be done automatedly: I would be pleased to know whether it can be so done or whether it cannot be so done.

I am thinking of trying to produce a font using this effect, a font with a name such as Ripples of Randomness.

I am also thinking of whether the effect of scaling so that effectively smaller random numbers are used could be used to produce an antique effect to a font.

William

[Bhikkhu Pesala]

Hi William

It is possible to stack the transformations so that they are done one after another:

1. Scale (400, 400) Centre
2. Random move points
3. Scale (25, 25) Centre
4. Left side bearing point at x=0

The same set of transformatons can also be applied repeatedly to roughen the font progressively until you're satisfied with the result.

[William]

Thank you for your reply.

I have just uploaded to the web the following font file.

Random 102

It is in a file RANDO102.TTF

It is available at the following web address.

http://www.users.globalnet.co.uk/~ngo/RANDO102.TTF

The font is not at this time listed on any of my web pages.

This font has only five characters in it. Each has lots of off-curve points on straight lines, ready for their positions to be randomly moved by the random point movement transform.

I started with RANDO101.TTF which just had the h in it.

The idea is that the RANDO1xy.TTF fonts are the originals, then RANDOzxy.TTF is the result of the (z-1) attempt at using one random point movement transform on RANDO1xy.TTF. So, my first try of one random point movement on RANDO101.TTF is in RANDO201.TTF, my first try of one random point movement on RANDO102.TTF is in RANDO202.TTF and my second try of one random point movement on RANDO102.TTF is in RANDO302.TTF.

Naturally, anyone trying one random point movement transform on RANDO102.TTF will obtain a different result from my result, unless by an extremely unlikely coincidence or maybe that perhaps the random number generator starts with the same seed value every time.

The next stage for me will be to try to produce RANDO103.TTF by adding some more characters to a copy of RANDO102.TTF.

I am thinking that as various similar yet different fonts can be produced from copies of one basic font by different runs of the random point movement transform, it might be possible to produce a set of several such fonts with the idea that one is printed red, then another is printed orange on top and then a third is printed yellow on top, so that in a pdf an effect of yellow letters with orange and red bits round some of the edges is produced. Or maybe several shades of blue and green to produce a novel effect.

However, I found it interesting to have a go with this and hope that the availability of the RANDO102.TTF font will make it easier for readers to have a go with this technique.

----

Using several goes at the random point movement transform in sequence raises some interesting matters concerning the way that several random movements interact in a random walk sequence.

Although not about fontmaking, some readers might perhaps like to have a look at the following which I wrote some years ago which is still in our family webspace.

http://www.users.globalnet.co.uk/~ngo/rwai0001.htm

William