How to Prepare an Image for a Font Generator

The preparation rules overlap, but the starting material is different. Open MakeFont Image to Font when you have separated letter images, a scanned lettering alphabet, or a permitted photograph whose characters you want to map individually. Open MakeFont Handwriting to Font when you have filled a handwriting sheet and want that sheet to become an editable Font Profile.

A beautiful page is not automatically a useful font source. The generator needs enough visual evidence to separate one intended character from the next and enough context for you to verify the proposed mapping. The Unicode Consortium’s confusable-character guidance (opens in a new tab) documents why shapes such as O and 0 can be visually confused; a written target list lets you resolve that ambiguity instead of asking image recognition to guess. These five checks catch most avoidable input problems before processing begins.

An image to font workflow is different from a font finder. A font finder tries to name an existing typeface that resembles a photograph. A font generator uses the shapes you supplied to create an editable character set. That makes missing edges, accidental marks, and uncertain labels part of the construction problem rather than harmless photographic noise.

The image first has to be divided into useful regions. Each region then needs a character label and a vector outline. Microsoft’s OpenType glyph-processing overview (opens in a new tab) explains that a font maps character code points to glyph IDs through its cmap data. After mapping, the outline still needs practical font information: bounds, a baseline, advance width, and enough coverage for the text you intend to type. A clean source does not make those decisions disappear, but it makes every decision easier to see and correct.

This is also why an imperfect but well-organized handwriting sheet can be more useful than a polished poster. The sheet tells you where one character begins, what it is supposed to mean, and whether important strokes extend above or below the writing line. The poster may look better while providing less reliable construction evidence.

1. 01

   Choose the character target first

   Write down the uppercase letters, lowercase letters, numbers, punctuation, or limited display set you actually need. A finite target makes missing characters visible and prevents a partial source from being mistaken for a complete font. If you are filling a character sheet, follow the Handwriting to Font workflow rather than treating the sheet as unrelated artwork.

2. 02

   Give every character its own space

   Use a grid, guide sheet, or generous rows. Keep neighboring strokes apart. Leave extra room below g, j, p, q, and y, and above capitals, accents, and tall ascenders.

3. 03

   Keep the writing conditions consistent

   Use the same pen, stroke weight, paper, scale, and general writing angle across the set. Intentional variation can be part of the design; accidental variation makes the first font harder to evaluate.

4. 04

   Capture the page straight on

   Prefer a scan when available. For a photograph, keep the camera parallel to the page, use even light, avoid a hand or phone shadow, and make sure the smallest punctuation remains in focus. The US Federal Agencies Digital Guidelines Initiative publishes broader still-image digitization guidance (opens in a new tab) covering capture quality and consistent imaging practice.

5. 05

   Crop around the set, not through it

   Remove unrelated desk or page area, but keep breathing room around the outermost characters. Check that no swash, accent, serif, or descender touches the crop edge.

6. 06

   Inspect the full-size image

   Zoom to 100% before upload. Look for motion blur, broken faint strokes, paper texture, compression blocks, and small marks that could be mistaken for punctuation. Recapture when the defect affects many characters.

7. 07

   Keep the untouched original

   Save a master scan or photograph before aggressive thresholding, sharpening, or background removal. If an edit erases a thin stroke, the original gives you a trustworthy way back.

MakeFont accepts PNG, JPG, and WEBP source images up to 20 MB. Inspection, OCR proposals, tracing, and editing run in the browser. The source image is kept in IndexedDB for the local project and is not uploaded or synced. Recognition remains a proposal: you review the character label, region, outline, bounds, baseline, and coverage instead of accepting a hidden automatic decision. The later build produces TTF and a web preview using WOFF2; the W3C WOFF2 Recommendation (opens in a new tab) defines WOFF2 as a compressed font format for Web use, not as a replacement for reviewing the underlying font data.

1. Choose the source locally and inspect the proposed regions.
2. Correct uncertain character labels before treating the mapping as final.
3. Edit bounds, baseline, outline, or strokes where the visual result needs correction.
4. Review found, uncertain, failed, and missing coverage as separate states.
5. When the Image to Font profile is ready, sign in to start the TTF/WOFF2 build and preview, then download the TTF. Return to the Image to Font workspace for targeted corrections, and see the current plan information before deciding whether the signed-in workflow fits the project.

These problems are useful signals. When the same defect appears across the whole sheet, improve the source. When one or two glyphs are wrong, keep the better source and make targeted edits. That distinction prevents a local correction from turning into unnecessary full-page rework.

• Missing characters: a clearer image cannot supply a letter, number, or symbol that was never included.
• Conflicting styles: if half the source is heavy marker and half is light pencil, better lighting will not turn them into one coherent design.
• Font identification: MakeFont creates an editable profile from supplied letterforms; it does not identify the name of an existing photographed typeface.
• Permission and originality: image quality does not establish copyright, trademark, licensing, or sharing rights.
• Unsupported writing systems: the first quality-gated targets cover selected Latin, Greek, and Cyrillic sets. Complex shaping systems remain outside the current release.
• Professional type design: clean tracing is a starting point. Kerning, spacing, language coverage, optical correction, and production testing still require judgment.

Only process material you are allowed to use, and review the acceptable-use boundaries before sharing or distributing a result. For details about local source handling and optional account features, read the privacy policy.

Do not judge a font only from an alphabet grid. Letters that look balanced alone can collide, float, or create uneven rhythm in real text. The OpenType horizontal metrics specification (opens in a new tab) distinguishes glyph outlines from advance widths and side bearings, which is why a recognizable outline can still space badly. Start with the exact characters your source covers, then test repeated pairs, short words, mixed uppercase and lowercase, numbers, and punctuation. Compare the preview with the source rather than assuming a technically valid file is visually finished.

1. 01

   Check the baseline

   Type words containing round, flat, tall, and descending letters. Look for characters that visibly float or sink.

2. 02

   Check side spacing

   Repeat combinations such as HH, OO, nn, oo, AV, To, and punctuation pairs that exist in your target. Uneven rhythm often comes from metrics, not the outline.

3. 03

   Check small sizes

   Thin gaps and interior counters that look open in a large glyph view may close in normal reading sizes.

4. 04

   Check coverage honestly

   Treat missing and uncertain positions as unfinished or intentionally partial. Do not describe a limited display set as a complete alphabet.

5. 05

   Keep the revision path

   Return to the editable Font Profile for targeted corrections, rebuild, and compare again instead of losing the link between source, glyph, and output.

You can inspect the current narrow build and its explicit evidence boundary on the MakeFont examples page.

• I know which characters this source is meant to contain.
• Every character has visible space on all four sides.
• Accents, punctuation, ascenders, and descenders are not clipped.
• The marks are in focus and clearly separated from the background.
• The page was scanned or photographed straight on with even light.
• No unrelated texture, ruled line, or shadow is likely to become part of a glyph.
• I kept the untouched original before applying image cleanup.
• I am prepared to review mappings, outlines, baselines, spacing, and missing coverage.
• I have permission to process and use the source material.

These sources explain the imaging and font-format concepts behind the practical advice. They do not endorse MakeFont and are not presented as product performance evidence.

• FADGI Technical Guidelines for Digitizing Cultural Heritage Materials (opens in a new tab) — federal still-image capture and quality guidance used here for the straight, evenly lit source recommendation.
• Microsoft OpenType glyph processing (opens in a new tab) — explains character-code-to-glyph mapping through cmap data.
• Microsoft OpenType horizontal metrics (opens in a new tab) — defines advance widths and side bearings used in horizontal text layout.
• Unicode Technical Standard #39 (opens in a new tab) — documents visually confusable characters and supports the recommendation to verify ambiguous mappings manually.
• W3C WOFF File Format 2.0 (opens in a new tab) — the current W3C Recommendation for the compressed Web font format used in browser previews.