You'll get different answers about the "right way", but to me best compromise is what I did in Patria:

1) Assuming you have a non-extreme weight at hand, use the software's "weight change" to make the light and dark extremes. Confine to the x-axis.
2) Tweak like there's no tomorrow, specifically by decreasing the contrast in the light extreme while increasing it in the dark. But you should also adjust the overshoots: the lighter the greater.
3) Create a new intermediate of where you started, and compare it to your original. Assuming you still like your original better, tweak the two extremes (not the resultant intermediate) to get back to where your original was.
4) Once you end up with extremes that you like, and they generate your original intermediate weight to your liking, you're done. You shouldn't have to tweak any intermediates at all. In general. :-)

Two words of caution:
1) Very light weights reveal problems in stroke thickness very obviously.
2) Very dark weights can get too pudgy. It takes a lot of expertise to put the sparkle in them - few people can manage - I for one can't. Look at the darkest weights of Galliard for guidance there.

BTW, you don't have to make an actual MM font - in fact most users would have trouble with that! Just have your axis "internal" to your font app, and generate the necessary intermediate weights. Four weights is a good number - too many weights and some users can get confused. If you have very light and dark weights, go to 6.

Spacing the weight gradation is v e r y t r i c k y ... and very open to personal preference.

hhp

Good tips, Hrant, but I avoid using automated "Change Weights".
I don't think it's worth it. A lot of tweaking will be necessary
anyway, so I just "change weights" myself, manually. It's like
splitting together your steps 1 and 2. ;-)

FontLab 4.5 ships with a demo Python script called 'Family Generator', which automatically produces a Light, Bold and Heavy design from a Regular. Of course, the results are never perfect and always require manual editing to clean them up, but depending on your design this can be surprisingly effective as a first step. The other thing to realise, of course, is that you can edit the Python script, and so tailor the weight and width adjustments made by the script to your design.

i think kyle is absolutely right. forget multiple master, it's hardly breathing anymore.

there probably isn't a correct way and it usually depends on the kind of design you're working on. i personally have found it to be efficient to work both ways at the same time. you can use the "change weight comand - resize to x-height" as a sort of reference and also to get your accents, punctuation etc. you still have to tweak those, but it's less of a hassle then also tweaking all the main letters from that. those you should definately draw manually.

it also helps layering all weights of one character atop of each other and optimizing the "new" weights right there in one window in comparisson with your finished regular weight.

"FontLab 4.5 ships with a demo Python script called 'Family Generator'"

I've searched and searched and can' find it..are you sure it came with FL4.5 or is it one of your own?

Yepp, but that's not included. Maybe something to pass along if it free?

Speaking of sharing scripts, does anybody here have RoboFog? There's a script by Peter van Blokland called "Corner Rounder" or something that I'd like to check out - not to copy or port, just to see if there's some smart code in there I could learn from.

hhp

There is a special mail list here:
http://mail.letterror.com/mailman/listinfo/fontlab-scripting

There is a Python page on our site: http://www.fontlab.com/html/python.html
but it is not updated very often.

You can download "Make Family" script from here:
http://www.font.to/_data/blend.py

It looks that for some reason we did not include it into the Mac package. We will fix that.

Best regards,
Yuri Yarmola

Yuri, thanks so much for the "Make Family" script. While it looks like I've got quite a bit of manual correcting ahead of me, it quickly produced a great base to start from!

Cheers!

Wow. Heavy. And a bit confusing still (to me).

So there's not the connection to Fontographer than RoboFog was based on, right?

hhp

One should consider that multiple master can be used as a design tool, without ever generating multiple master fonts in the end. Most recent Adobe fonts were designed in MM space, without ever being built as MMs.

Personally, I would add a weight axis to the regular, design the black weight, extrapolate the light weight, set it as the new master at the light end, and then clean it up.

One good piece of advice Robert Slimbach gave me on the family I'm working on right now was that designing the extreme masters and interpolating really helps the intermediate weights come out better. It forces you to fix things in the extreme masters that might not be so noticeable in intermediate weights.

Cheers,

T

Well, "should" is to strong a word. I'm saying you can do it, and that for anything more than three weights it can be faster (although maybe not the first time you do it). Plenty of very nice typefaces have been developed this way, including most of Robert Slimbach's designs.

There are cases where you wouldn't want to do it this way, or where you'd use an intermediate master on the weight axis (something that FontLab doesn't directly support).

Regards,

T

:sigh: So much to think about.

I get enough headaches as it is

> there is a lot of cleanup

Unless it's an overly complex design with an inherent issue with interpolation, I find that iteratively tweaking the extremes to get intermediates that don't need any adjustments themselves (except correcting rounding errors) is of great benefit. The reason isn't just design elegance, it's practicality: if you change your mind about what the intermediate(s) should be, or if a client wants a custom instance later on, it's much quicker.

hhp

I heard Jonathan Hoefler describe the development of
Champion in the same way Thomas describes. Of
course that was 4 years ago, so I'll have to fudge the
details. But in essence the method is to start with the
regular weight, then manually design the extremes (the
widest black and the narrowest thin). Then have the
software create the basis for the intermediate weights. (If
my memory is correct, the software would've generated 3
weights of Champion.) Then of course, there is a lot of
cleanup work to do to add some human care/finesse to the
intermediate weights.

In other words, the computer can help take some of the
tedium out of the process, but it can't make up for
attention to detail.