Font revival from A to B


Dear typophiles and experts,

I'm a humble amateur who knows next to nothing, but I do love typography like few other things in life. I'm kindly asking for your advice on an unusual question. 

I'm trying to learn how to get from a scan of an old face (say, 1200 dpi) to a FLS file. One I get a .vfb file in place, I'll know what to do. 

My intention is to make a homebrewed revival of a late 19th-century face for my own, purely personal use. I'm not a graphic designer, not a typographer, not a developer of anything. Just a guy who loves books and has to spend a lot of time looking at printed matter. I don't plan to share this font with anyone, let alone sell it. 

With that in mind, I'd like to ask for your opinion on this question: 

Where/How can I learn relatively easily to move from a bmp/png/svg file to a Fontlab-compatible file? 

Strong desiderata:

1.  Shallow learning curve. Regrettably, I have a day job and a family. I'd love to spend a lot of time learning to do this like a pro, but I can't afford that time and effort. 

2.  Intuitive explanations. I've spent the last 20 years wading through rivers of unhelpful, meandering, confusing talk by my peers. One thing I learned is: intuitive, clear explanations are the best. I don't mind being treated like a child. In this particular matter, I'm no better than a child. 

Thank you all for your attention and advice!

Comments

  • Well there's FontLab's Scanfont product. And FontSelf. But I haven't used them, and would recommend these basics while working in a conventional font editor:
    — Nail down the vertical guides within the EM before doing any actual work. You may need to add an artificial vertical reference bar alongside the letterform scans so that they import predictably (with no need to fudge the scaling by hand after the fact).
    — Don't auto-trace, trace by hand over the scans.
    — Place points on all extrema, and other points only if the shape truly demands it.
  • Scanfont is no longer offered, but there are a lot of options available now. 
    However, for much of the effort needed to create a worthwhile revival, the choice of software is probably a secondary matter. 
  • Thomas PhinneyThomas Phinney Posts: 2,732
    edited June 2021
    Newer versions of FontLab (6 and 7) have ScanFont-like functionality built-in. Even auto-separation with OCR, so as to put the scans into their correct cells.

    Even if you are like me and throw away the auto-trace, having the scans already sized and positioned and sliced up into their cells is extremely helpful.
  • Thank you all! This is already very useful. I had a hunch that FLS 7 already has most of what I need built into it, but it's good to have it confirmed by pros. 

    Other tips or links to tutorials welcome. 
  • Paul MillerPaul Miller Posts: 273
    High Logic Font Creator has all the functionality to import bitmaps and trace the outlines.  Then you will usually need to tidy them up a bit.
    They also have a product specially designed to do this job, but I have never used it.
  • Igor PetrovicIgor Petrovic Posts: 262
    edited June 2021
    I never was into a font revival venture, so I don't know how this is helpful and intuitive but I see that Frank Blokland posts on his Facebook page regularly on this matter. I also think that I saw a piece of software they developed for this purpose, you might want to check 

    https://www.facebook.com/lettermodeller 

    P.S. I've just scrolled his wall and found this website about the revival project, it might be useful for you:

    https://www.rosart.nl
  • Paul MillerPaul Miller Posts: 273
    My intention is to make a homebrewed revival of a late 19th-century face for my own, purely personal use. I'm not a graphic designer, not a typographer, not a developer of anything. Just a guy who loves books and has to spend a lot of time looking at printed matter. I don't plan to share this font with anyone, let alone sell it.
    The only revival I ever did was Munson, the photographs I was copying from were of low quality and incomplete in their coverage.  It is often better to start from scratch and construct a font which subjectively looks like the the font you are trying to revive.
    For instance on the photographs the corners of the serifs looked quite rounded but I thought that was just an artifact of the printing and the subsequent photography so I went with the square corners, after all it is a slab serif and that was the look I was after.
    In my (admittedly extremely limited) experience of revivals I would say, yes, scan the originals and see what you come out with but the scans are a starting point not an endpoint.  You may want to edit them to make them look like the font you subjectively perceive on those old pages as you think it was meant to be.
    Unless of course you want to include the rounding effect of being printed with viscous ink on rough paper.  ;) It's up to you.  Good luck.
  • My first outline font was a revival but I didn't have a scanner so I eyeballed it.  :-)  It came out even worse than the original.
  • Rob BarbaRob Barba Posts: 86
    High Logic Font Creator has all the functionality to import bitmaps and trace the outlines.  Then you will usually need to tidy them up a bit.
    They also have a product specially designed to do this job, but I have never used it.
    I'll second the High Logic angle. That's my main program and it does exactly as it says with very little fuss.

    I'll played around with a copy of Scanahand, but that seems more for pure amateurs, which you seem to be a little above, so I would recommend the FC angle.

  • konrad ritterkonrad ritter Posts: 200
    edited June 2021
    Thank you, Rob and everybody. Much appreciated. Between their FontCreator and the new FLS 7, I guess I'll be OK. 
  • Igor FreibergerIgor Freiberger Posts: 250
    edited June 2021
    Konrad, this sequence of instructions from our user manual may help you:

    From the serie of video tutorials recently published, you may see

    You can also ask for additional instructions in our forum or directly through support requests.
  • This SUPER helpful, and exactly what I was looking for. Tremendously good advice, and right on the money. So grateful for your help, Mr Freiberger. Thank you much!
  • konrad ritterkonrad ritter Posts: 200
    edited June 2021
    Thank you, Mr Hudson. Your question goes to the heart of the matter, but that makes it the hardest to answer. 

    The only answer I relies on metaphors, which I hate in this context; here literal language does best, and imagery is of little help.

    I know what I'm looking for -- it's coming close to a "what it feels like" sort of thing. Dredging up a digital simulacrum that'll help me feel closer to the experience I had as a young man looking at those printed pages, turning them, and glancing at the whole page out of focus, letting the particular texture of that type burn itself into my retinae, disregarding completely what the particular words were trying to say. 

    Two more attempts to say what I'm looking for. One: I grew up reading a lot of French prose printed in metal cuts of Series 16eme and 17eme (see also here: https://typedrawers.com/discussion/3655/reviving-french-classics-from-the-later-lead-age#latest). For me, those two faces have a feel. None of the digital revivals of Serie 16eme, from Berthold's Augustea to Charles Maze's Berthe, comes close to recreating that feel for me. However, Mr Simonson's Etna does -- inadvertently, as it were, but it succeeds admirably. When I look at it, I feel transported to the place I yearn for. 

    Another analogy: when I read things set in your face, Brill, I feel the same vibe I get from reading books from 18th-century British presses. It's a feel I never get from the many digitizations of Baskerville I've seen. 

    At the same time, my original query was a lot more humble. I was just looking for help getting from A to B. Many of the answers I got regard the journey from B to Z. That takes years of learning and struggling for wisdom. Who knows where that journey will take any of us who tries it. 
  • I have one piece of advice before starting work on a typeface revival: sit down and form a mental image of what you want a page of text set in the resulting type to look like. This will help you determine the approach to take to the revival: do you want a slavish reproduction of the metal type as used in a particular book or specimen? or do you want a more rationalised interpretation? or do you want to use the structure of the original to carry a new character, perhaps sharper or softer? Having this in your head before you start will save you a lot of time fiddling with scans and traced outlines trying to find your way without a destination in mind.
    Sure, that's the most important question. Roughly I see 3 levels:

    1. Reconstruction: an attempt to reconstruct the shape, look and feel and even the spacing, making it possible to render a page of an old book without the technical failures (broken or deformed letters, over-inking, wavy baselines). But it needs e. g. more or less rounded corners to imitate the squash of the ink at the edges.

    2. Revival: Keep the shape and proportions (more or less) with some optimisations for modern use.

    3. Redesign: Take the original as an inspiration and put own personality it.

    In most cases some characters will be missing in the specimens and need creation.

    Just for the records an example of too slavish:



    The above picture is from a specimen of "Original-Breitkopf-Fraktur" recut by H. Berthold AG at the beginning of the 20th century. Take a closer look at \c, \c_h and \c_k. Some of the digital versions copied this bad designed \c_h which is far away from the originals by Breitkopf jun.
  • Although as with any act of creating, let the thing speak back to you about what it needs to be.
  • Helmut’s categories remind me of a seminal text on this topic: John Downer’s “Call It What It Is”.
  • 1. Reconstruction: an attempt to reconstruct the shape, look and feel and even the spacing, making it possible to render a page of an old book without the technical failures (broken or deformed letters, over-inking, wavy baselines). But it needs e. g. more or less rounded corners to imitate the squash of the ink at the edges.

    2. Revival: Keep the shape and proportions (more or less) with some optimisations for modern use.

    3. Redesign: Take the original as an inspiration and put own personality it.

    In most cases some characters will be missing in the specimens and need creation.
    This is interesting because when I decided to tackle De Vinne, before any of my own designs, I considered the thing from many angles. I did not re-read John Downer’s essay, I did not want to attain 1., even less 3.
    But in 2. there is a lot of room about what these "optimisations" could be.

    Personally I tried (and I am finishing the first three styles right now) to *capture* what I felt as the essence of the curves. I considered Optical size, I tried to base my initial version on a 72pt size. Since not all glyphs were obviously available at the given size, I kept collecting the most varied examples (Jacob Casal even photographed for me a good set of a 30pt size Roman).
    Recently George Thomas was so nice to make macro-photos (as scans) of De Vinne Roman (the text version by Goudy) for me. By working on De Vinne alone for months, this really helped me to get into the curves, so when I added accents, new glyphs and new typographic elements (be them period pieces or new ones which nonetheless sit well with the design) I always tried the exercise to imagine how De Vinne would have looked if produced now, of course from hand drawn letters. I find the result satisfying as it looks precise on one side — some elements I introduced even purposefully “stride" with the original forms, like the Math symbols which mostly were not there — while keeping all the inconsistencies and irregularities that are exquisitely representative of the precise time span when it came out.

    To cut a long story short I would ask: would you be satisfied with autotraced bezier curves (even if "polished" and maybe hinted)? I understand you’d have very little time to properly learn to draw/finalize in vector forms, but I know that on my part I would not be satisfied.
  • Brief answer: yes. I would be satisfied, at least for a while. Over the years, I might try to take it into a new direction. Again, this is not a commercial project. Plus, I learned that we get wiser over the years -- so, I'm sure that later down the road, I'll see better ways to tweak a font. For now, I just plan to start small.  
  • Thomas PhinneyThomas Phinney Posts: 2,732
    I have never been happy with the un-tweaked results of autotracing, unless the purpose was to make a deliberately rough and uneven typeface mimicking uneven/worn metal type.
  • I have never been happy with the un-tweaked results of autotracing, unless the purpose was to make a deliberately rough and uneven typeface mimicking uneven/worn metal type.
    If the purpose is that, I found autotracing equally bad, as the results are qualitatively poor in terms of "simulation".
  • Nick CurtisNick Curtis Posts: 118
    I draw original outlines for all of my revivals / reconstructions / redesigns using orthogonal ellipses / arcs, and I concur  that auto-tracing is good only for deliberately crappy contours.
  • I also think that I saw a piece of software they developed for this purpose...
    I guess you are thinking of Glyph Collector? It is a piece of software that makes an average shape based on several appearances of the same letters within a few scanned pages. A good starting point imho.
  • I guess you are thinking of Glyph Collector? It is a piece of software that makes an average shape based on several appearances of the same letters within a few scanned pages. A good starting point imho.
    AFAIK Glyph Collector doesn't calculate an average shape. It helps collecting glyphs and the user selects one.

    There are other scientific projects in the context of OCR correction, that generate fonts like Terese. It overlays the generated font to make differences visible for proof reading:



    Others calculate special images of glyph prototypes convenient for similarity calculation. None of them can generate a nice fonts. I use similarity hashes (fingerprints), which I can generate from fonts, printed specimens or scanned books.

    Others mainly in East-Asia focus on font generation. Either tracing or applying a style to skeletons. This saves time in case of the large number of Chinese glyphs (Han, Kanji).

    They use special algorithms for tracing (better than autotrace or potrace) and scan from large paintings or drawings (US-letter/A4 for each glyph?).

    Even a scan at 300 ppi of letters printed at 60 points results in ~280 pixels height. With potrace it looks like the scan with all the speckles (also white speckles). Correcting this and the edges, corners etc. in a glyph editor takes much time. It results in the experience, that using a scan in the background layer is a better way. Believe the experienced font drawers. Been there, done that.

     
  • AFAIK Glyph Collector doesn't calculate an average shape. It helps collecting glyphs and the user selects one. 
    It was my impression that's exactly Glyph Collector's strength. To quote @LeMo aka PatternMan aka Frank E Blokland in https://typedrawers.com/discussion/977/glyph-collector
    The tool is meant for collecting multiple representations of glyphs from a scanned page, and for subsequently generating an average image.
  • OK, now I remember and retried Glyph Collector.

    You need a scan of e. g. a page. You also need an image file of each character. This means, open the scanned page in an image program like Photoshop, cut out the character and save it to a file. I used Gimp and needed 10 keystrokes for each character. Did only a-n for a test.

    Next import the images of the single characters into Glyph Collector and run them against the scanned page(s). It collects all images of the same character. It should.

    Here is the result of \i which appears 10 times in the text:



    It detected 3 (one of them is another size and proportion) and 8 false positives.

    Now you must throw away the wrong ones, mark the good ones and click "Generate Average". It really just does an average. Now open your font editor and import the average images for each glyph into a background layer.

    Or autotrace them by a script and import the SVGs.

    You still need to scale and align each glyph. No help about descender, ascender, spacing, kerning/overlapping. This information is lost in the process.


Sign In or Register to comment.