Desenho De Fontes Históricas

On redoing type from old books for the third time this month

I was digitizing a 15th-century Portuguese manuscript letterform set last week and ran into a problem most people don't warn you about. The scribe used a quill, which means the stroke width varies constantly depending on pressure and angle. When you try to trace those letters directly in Illustrator or FontForge, you end up with a font that looks like every character was drawn with a different pen, which defeats the purpose of creating something usable. The workaround I settled on is to sample multiple instances of each letter across the manuscript, find the common denominator in stroke thickness and proportions, and then draw a clean vector from that average rather than copying any single example.

What desenho de fontes históricas actually involves

It's not just tracing. Anyone can trace an old letter. Real trabalho here means understanding why the letter looks the way it does, then rebuilding it as a functional typeface. You're studying primary sources -- incunabula, stone inscriptions, illuminated manuscripts, wax tablets, woodcuts -- and extracting the letterforms from them. The output should be something you can actually use at multiple sizes, not a museum-grade reproduction that falls apart at 12pt. The tools most people use are FontForge for the open-source route, Glyphs for Mac, or Illustrator if you want to hand-draw and then import. FontForge handles the math for Unicode encoding and kerning pairs. Illustrator or Affinity Designer is better for the initial drawing because the bezier control is smoother. I keep both in my workflow.

Where to find sources you can actually work from

Most historical sources live behind paywalls or in poorly digitized archives. The sites worth knowing are the Digital Scriptorium, Gallica from the BnF, the Internet Archive's incunable collection, and the British Library's digitized manuscripts. The Cambridge Digital Library is also excellent for medieval materials. For Portuguese specifically, the Torre do Tombo digital collection and the Biblioteca Nacional de Portugal have high-resolution scans that are actually usable. Download these as large TIFF or JPEG files, not PDF. PDFs compress the images badly and you'll waste hours cleaning up artifacts in post. If the source is on parchment or vellum, expect discoloration and wormholes. These don't affect the letterforms much if you're looking at the right resolution, but they make automated tracing tools completely unreliable. I've tried the auto-trace features in Illustrator and FontForge on degraded manuscripts. They produce garbage. Hand-draw every character.

The actual workflow I use

First I print the source at 1:1 scale if possible, or at minimum 200 percent. Drawing on screen at full resolution feels wrong because your hand doesn't have the same control. I use a Wacom tablet connected to Illustrator for the initial drawing phase, then move to FontForge for encoding and metrics. The whole process for a complete alphabet with accents runs about 40 to 60 hours for a competent person. Beginners take longer because they obsess over individual glyphs instead of establishing consistent proportions first. Start with the uppercase, then the lowercase, then numbers, then punctuation. Handle the accented characters last. That order matters because the unaccented base forms inform the accented versions. If you do accents first, you'll redesign the base later and everything breaks.

Establish your x-height, cap height, and em square early. For historical work, these proportions often differ from modern defaults. A Gothic textura might have an x-height that's nearly equal to the cap height. A humanist italic will have a much smaller x-height relative to caps. Match the source, don't force it into a modern grid.

Problems you will hit and how to fix them

The biggest issue is inconsistency in the original source. Medieval scribes changed their handwriting within a single document. A well-trained eye catches this; a beginner copies everything literally and ends up with a font where "e" appears in twelve different shapes. Your job is to identify the dominant form and make decisions about the outliers. I usually keep a reference sheet showing every variant I encountered, pick the most representative one, and note the exceptions in a design log. Another common problem is that historical fonts weren't designed for digital use. They have terminal flourishes, variable stroke weights, and decorative elements that look terrible at small sizes. I solve this by creating two versions: a faithful reproduction for display purposes and a simplified alt set for body text. FontForge lets you map alternate glyphs through the glyph alternate feature, so users can choose which version they want.

I encountered a specific case last year where I was working from a 16th-century Spanish woodcut typeface. The source had prominent ink bleed effects because woodcut printing works differently than metal type. The bleed made the letterforms look thicker and slightly irregular. If I reproduced that exactly, the font would be unreadable at any size below 18pt. I traced the actual carved letter areas, not the ink spread, and then adjusted the weights to be slightly lighter than the original. The result looked authentic but was actually more legible than the historical source itself, which is the right call for a functional typeface.

Encoding and Unicode considerations

Don't skip this. A beautiful font with no Unicode mapping is useless in modern workflows. Map your glyphs to the appropriate Unicode blocks. Latin Extended-A covers most European historical needs. Latin Extended-B handles some older characters. If your source uses long-s or swash capitals, you need the proper Unicode slots for those, not just custom alternates. FontForge's encoding editor makes this straightforward once you understand the blocks. Kerning is non-negotiable for historical fonts. These typefaces often have tighter optical spacing than modern designs. I spend more time on kerning pairs than on individual glyph drawing. Start with the obvious problematic pairs -- AV, To, Te, Ly, Pa -- and work outward. The Kern pair editor in FontForge isn't pretty but it gets the job done. Glyphs has a better interface if you're on Mac.

When this approach fails

If your source material is extremely damaged or sparse, handmade tracing won't save you. I've worked from manuscripts where entire alphabets were missing or only fragments survived. In those cases, the honest approach is to document what exists and leave the rest as design decisions based on comparative analysis of contemporary sources, not to invent characters that have no historical basis. I've seen people fill gaps by borrowing from neighboring scripts or guessing at missing forms. That's not reconstruction, that's fabrication, and it damages the credibility of the entire project. Also, historical font drawing is computationally heavier than modern type design because each glyph often requires more manual adjustment. Expect your rendering times to be longer and your file sizes to be bigger. A well-made historical font can easily exceed 15MB with full kerning and alternate sets. If you're distributing this online, consider providing a subset version.

Software recommendations

FontForge is free and handles everything end-to-end. Glyphs is better for Mac users who want a polished interface. RoboFont is another solid option but steeper learning curve. For the initial drawing phase, Illustrator or even Inkscape if you're on a budget. I've used Inkscape for historical work and it produced acceptable results, though the bezier handling isn't as smooth as paid tools. There's also Calligraphr if you want to skip vector drawing entirely and work from handwritten samples, but that approach sacrifices precision for speed. If you want a quick reference to get started, there are template documents available on GitHub that show standard metric grids for various historical periods. Searching for "historical typeface metric templates" will surface a few useful ones. They won't replace understanding the source material, but they'll save you from starting with arbitrary proportions.