10 fonts tested: coverage, not ink saving promises
Startoner · Digital study · 2 October 2026
At 12 points, Liberation Serif leaves 24.7% less black area than Liberation Sans in our document. When equalising lower-case height, the ranking changes: the visible size matters.
Original conceptual illustration generated with AI. Does not represent measured pages or physical print testing.
Choosing a font to save ink seems straightforward: pick the one that leaves less pigment on the page. The problem arises when one typeface appears visibly smaller than another even though both share the same nominal point size. To investigate this, we compared ten font families using the same Spanish test text and repeated the calculation after equalising the x-height of their lowercase letters. The result is a fully reproducible table, complete with data files and explicit methodological boundaries.
Free download
What it means for a font to cover less paper
Digital coverage measures the raster pattern produced by the document file. Here, we calculate the sum of all pixel blackness values across every page. A pure white pixel contributes zero, a pure black pixel contributes one, and antialiased edge pixels contribute a fractional value. This prevents smoothed font contours from being counted as solid black ink. The total sum is then converted into equivalent surface area in square millimetres.
This calculation reflects the chosen rasterisation engine. It does not measure optical density, ink volume, or toner mass. It evaluates digital grayscale values without physical reflectance correction. Linking these numbers to actual printer consumption requires dedicated physical hardware experiments. As an example of printer consumption unrelated to printed letterforms, HP documents a maintenance cartridge that collects waste ink [1]. That device is not our test printer: no physical printer was used in this study.
Ten fonts, one text and identical conditions
We selected ten font families locally available in LibreOffice on the test workstation: Liberation Serif, Liberation Sans, Liberation Mono, Noto Sans, Noto Serif, Carlito, Caladea, DejaVu Sans, DejaVu Serif, and Rubik. Each was tested in its Regular weight (or Book for DejaVu), with no artificial bolding applied. These represent ten distinct families, not ten weights of a single typeface. The selection reflects local availability and does not claim to represent all typographic designs.
The files, versions, hashes, and original paths are fully documented. Font licensing was verified directly: Liberation, Noto, Carlito, and Rubik declare SIL OFL; DejaVu retains its Bitstream Vera-derived licence; and our Caladea 1.002 declares Apache 2.0. Copyright notices and licence texts are included in the bundle. We do not retroactively assign current licences to older software builds.
The document contains 695 words, counted by splitting on spaces, in eight original paragraphs. It is Spanish prose with accents, eñes, opening punctuation marks and some fictitious figures. All content is identical in every PDF. The layout is A4 portrait, with 20 mm margins, an additional horizontal allowance of 2 points on each side, fixed line spacing of 16 points and 8 points between paragraphs. The first comparison uses 12-point type, left alignment and pure black.
Each typeface was allowed to break lines naturally according to its glyph widths. We did not force uniform line breaks or trim sentences to fit. The rendering pipeline did not apply kerning or open-type ligatures, which represents a deliberate limitation compared to professional typesetting engines. No headers, footers, page numbers, or labels were added to the measured PDFs to avoid skewing pixel counts. The results belong strictly to this layout and cannot be extrapolated to unconstrained exports from Word or LibreOffice.
Results at 12 points: ranking by total area
| Rank | Font | Total black area, mm² | Average coverage | Pages | x-height, pt |
|---|---|---|---|---|---|
| 1 | Liberation Serif | 5,344.41 | 4.28 % | 2 | 5.508 |
| 2 | Carlito | 5,889.17 | 4.72 % | 2 | 5.730 |
| 3 | Caladea | 5,900.25 | 4.73 % | 2 | 5.604 |
| 4 | Liberation Sans | 7,097.04 | 5.69 % | 2 | 6.340 |
| 5 | Noto Sans | 7,201.11 | 5.77 % | 2 | 6.432 |
| 6 | Liberation Mono | 7,449.86 | 5.97 % | 2 | 6.340 |
| 7 | Noto Serif | 7,522.47 | 6.03 % | 2 | 6.432 |
| 8 | DejaVu Serif | 7,562.36 | 6.06 % | 2 | 6.229 |
| 9 | Rubik | 7,705.21 | 6.18 % | 2 | 6.240 |
| 10 | DejaVu Sans | 7,824.56 | 6.27 % | 2 | 6.562 |
Source: own measurements, 12pt and 300 dpi rows. The percentage represents equivalent black area divided by total page pixels across all pages. Ranking is based on absolute total document area, not average page coverage. Each font name links to its corresponding PDF.
Liberation Serif recorded 5,344.41 mm² of equivalent black area, compared to 7,824.56 mm² for DejaVu Sans. The relative difference between the two is 31.7%. This indicates lower digital coverage for this specific text under this conversion pipeline. It does not imply a 31.7% cost saving on replacement cartridges. Liberation Sans serves as an explicit baseline within the set, without treating it as an industry-wide universal standard.
It is also worth looking at page count and x-height. A low percentage can be misleading if a document uses more sheets and leaves a lot of white space. To avoid this, we rank by the absolute sum of black area. In this corpus, each of the twenty versions occupies two pages, although line breaks and the portion of the second page used vary. The page-by-page log allows that sum and its denominators to be checked. A monospaced font can also change the number of lines: its result does not make that type of font a poor choice for code or aligned tables.
Normalising x-height: the comparison changes
Typographic point size does not ensure equal visible letter height. Lowercase x-height defines the visual scale of lowercase characters, as detailed in the OpenType specification [2]. Thomas Phinney highlighted this critical variable when analysing ink-saving claims based on changing fonts [3]. Comparing fonts solely by point size inherently favours typefaces designed with smaller proportions.
In our second test run, we normalised the x-height of each font family to exactly 6 points based on measured font glyph outlines. The required point size was calculated from internal font design units without pre-rounding. We measured actual glyph outlines rather than relying on optional OS/2 header fields. Body text, line spacing, margins, and composition parameters remained constant. Resulting point sizes are rounded for display in the table; exact figures appear in the CSV.
| Rank | Font | Adjusted size, pt | Black area, mm² | Pages |
|---|---|---|---|---|
| 1 | Noto Sans | 11.194 | 6,261.37 | 2 |
| 2 | Liberation Serif | 13.072 | 6,341.98 | 2 |
| 3 | Liberation Sans | 11.357 | 6,353.94 | 2 |
| 4 | Carlito | 12.564 | 6,454.40 | 2 |
| 5 | Noto Serif | 11.194 | 6,548.92 | 2 |
| 6 | DejaVu Sans | 10.971 | 6,554.99 | 2 |
| 7 | Liberation Mono | 11.357 | 6,667.62 | 2 |
| 8 | Caladea | 12.848 | 6,766.01 | 2 |
| 9 | DejaVu Serif | 11.560 | 7,020.92 | 2 |
| 10 | Rubik | 11.538 | 7,132.68 | 2 |
Source: own measurements, xheight and 300 dpi rows. Common x-height: 6 points. This adjustment does not automatically equate readability.
Noto Sans now leads the ranking. Liberation Serif moves from position 1 to position 2. This change is the useful finding: choosing a winner without stating which size has been equalised can conceal part of the comparison. Equalising x-height does not equalise widths, spacing, stroke contrast, apertures or character shapes either. We have not measured reading speed, comprehension or accessibility; therefore, neither table establishes that all ten options are equally comfortable to read.
How we calculated and verified the numbers
We generated the PDFs with ReportLab and embedded the fonts. We converted them with Poppler into grayscale images at 300 dots per inch, with antialiasing enabled. We repeated the conversion at 600 dots per inch to check sensitivity to resolution. The maximum absolute variation in area between the two conversions was 0.31%; the rankings were not identical in every case, so both resolutions must be consulted. Under the equalised x-height condition, Noto Serif and DejaVu Sans exchange fifth and sixth places when the resolution changes; such a small difference does not support a firm recommendation between them. This comparison is neither a confidence interval nor a validation of physical consumption.
For each document, we verified that extracted text matched the source prose after whitespace normalisation. We also confirmed that all glyphs exist in the font files, contour bounds stay within page margins, and no lines overlap vertically. Raster auditing confirmed that no stray pixels fall outside the printable area. Full logs, line break data, and environment specifications are preserved in the methodological appendix · documentation in Spanish (ZIP).
The dataset is based on a single sample text created for this benchmark. Different languages, capitalized headings, tables, or bold styles will alter results. A 'Regular' font designation does not guarantee identical visual weight across different type design traditions. Font hinting and rasterisation engines also influence pixel generation. Published figures are stated to two decimal places for auditability, not to imply physical precision beyond the scope of digital image analysis.
How to apply the results to your documents
Begin with a representative excerpt of your own document. Change the font while preserving font size and inspect lowercase legibility. Check page count, accent rendering, numerals, and page breaks. If you find yourself increasing font size to read comfortably, accept that adjustment before making comparisons. Forcing a smaller font that strains the reader provides no practical benefit.
Use these benchmark tables to select font candidates, not to make savings claims. You can download the PDFs and evaluate them at actual print size; our real preview shows a sample page from the study. To reduce paper usage, consider double-sided printing. Consumer organisation OCU has warned that fonts designed to save ink can inadvertently increase paper consumption [4], making it essential to evaluate paper and ink decisions separately.
What about cost per page?
This study does not convert square millimetres of black area into currency. Estimating printing expenses requires actual consumable specifications and real usage data, as explained in our guide on how to estimate printing costs. It is equally important to understand what declared cartridge yields represent and distinguish lab ratings from your everyday documents. Our guide to office cost per page walks through those calculations.
If you need replacement supplies, you can explore compatible ink cartridges at Startoner and verify the correct reference for your printer model. We have not used retail prices or commercial page yield figures to fabricate artificial typographic savings claims. The contribution of this benchmark is transparent: an open dataset of pages, fonts, and calculations allowing anyone to examine what changes when selecting a different typeface.
Sources and references
Was this article helpful?
Ask us about your printer
Tell us the model and we will confirm the right cartridge. We reply within 24h.