Kerning Pair Inspector
Pair
Pair rendered
AVAV
Left: kerning enabled. Right: kerning disabled — any visible difference is the font's own pair adjustment.
Pair table
| Pair | Shaped | Sum of parts | Δ | Em/1000 | Verdict |
|---|---|---|---|---|---|
| AV | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| To | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| Yo | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| LT | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| Wa | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| P. | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| F, | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| rg | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| 11 | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
| fr | 0.00px | 0.00px | +0.00px | 0 / 1000 em | no kerning applied |
Measurements come from the browser's shaping engine at the size above; the em/1000 column is the unit kerning tables are authored in.
Reading kern values
- · A negative Δ means the pair was tightened — the usual case for diagonals meeting verticals (AV, To).
- · −20 to −60 em/1000 is a normal kern; beyond −100 the pair is visually merged and often needs optical spacing instead.
- · A zero Δ does not prove the face lacks kern data: many pairs genuinely need no adjustment.
- · Tracking (letter-spacing) is added on top of kerning and will hide small adjustments — measure with tracking at 0.
.kern-pair {
font-family: 'Fraunces', serif;
font-size: 4.0000rem;
font-kerning: normal; /* required: none disables the pair adjustments below */
}You Might Also Need
The kerning pair inspector measures the difference between a shaped pair and the sum of its two glyphs, which is exactly the adjustment the font applied. The number comes from the browser's own shaping engine rather than from a table copied out of the file.
What is Kerning Pair Inspector?
The kerning pair inspector compares two measurements of the same pair: the width of the pair as the browser shapes it, and the sum of the widths of its glyphs measured separately. The difference is the kerning adjustment the font actually applied, read from the renderer rather than from a specification.
The inputs are deliberately few, and two of them matter for accuracy.
- Face, from the catalogue, since kerning data varies from family to family and even between weights.
- The pair to inspect, from the classic A and V through to lower-case pairs and figure pairs.
- The measurement size, from 12 to 200 pixels, which should be large enough that the renderer does not round the result.
- A pair list, space separated, which produces a table of comparisons in one pass; the current pair is always included.
The readouts give the shaped width, the sum of the glyph widths, the difference in pixels and the same difference in em/1000 units, which is the unit kerning tables are authored in. The verdict line turns the number into a reading: a negative difference means the pair was tightened, which is the usual case where a diagonal meets a vertical, as in A and V or T and o; a positive difference means the font adds space; and a difference of zero means no adjustment was applied, which does not prove the face lacks kerning data, because plenty of pairs genuinely need none. The panel's guidance puts a normal kern between twenty and sixty units, notes that beyond a hundred the two letters visually merge and optical spacing is the better fix, and warns that letter-spacing is added on top of kerning and will hide small adjustments, so measurements should be taken with tracking at zero.
The rendered preview makes the comparison physical: the same pair is drawn twice, once with kerning on and once with kerning disabled, and the difference between the two columns is the font's own adjustment. The exported CSS is a small rule whose only interesting line is the kerning property set to normal, with a comment explaining that leaving it unset or disabling it discards the pair adjustments the measurement just quantified.
Four limits apply to what the measurement can claim.
- The measurement describes the browser's shaping at the chosen size and with default features, so a pair that changes under a discretionary ligature or a stylistic set will not match the number here.
- The sum of the parts assumes each glyph is independent, which is true for two letters and not for a sequence containing marks or combined forms.
- The rendered on-off comparison depends on the platform having a fallback face to compare against, so the two columns can look identical on a system with unusual font matching.
- The thresholds in the verdict column are the tool's own reading of accepted practice rather than a specification, useful as a scale and not as a rule.
What it settles reliably is the common case: two letters, one face, a size large enough to avoid rounding, and a figure in units per thousand of the em that can be compared with any other pair in the same face.
How to use Kerning Pair Inspector
- Choose the face first, because kerning data differs per family and per weight and the measurement is only meaningful for the face in use.
- Enter the pair and raise the measurement size above about 48 pixels, so renderer rounding is small relative to the adjustment.
- Read the difference in em per thousand units, which is the figure to compare against a normal kern of twenty to sixty.
- Compare the two rendered columns to see whether the adjustment is visible, then copy the rule with kerning explicitly enabled.
When to use Kerning Pair Inspector vs related tools
Reach for it when text set in caps looks unevenly spaced, when a display setting has two letters that appear to collide, or when someone claims a face ships no kerning and the claim needs testing rather than repeating. The rendered on-off comparison is the proof. For spacing across a whole run the letter spacing calculator and the font letter spacing tester handle tracking, the OpenType feature explorer shows the feature tags that change shaping, and the glyph comparison tool sets the same pair in several faces at once.
Privacy & Security
This tool runs entirely in your browser — no data ever leaves your device. There is no server round-trip, no upload, no logging, and no account required. Your input is processed locally using client-side JavaScript and is never stored, transmitted, or accessible to anyone else. When you close the tab, everything disappears.