常見問題

How to find a font from an image — step by step

Every designer knows the moment: a poster, a package or a website uses a typeface you love, and nothing in your font menu looks like it. Typing “what the font is this” into a search engine returns forums full of guesses. The faster route is to find the font from an image directly — here is how that works on this page, from screenshot to answer.

1. Capture a clean sample. Screenshot the text or photograph it as straight-on as you can. A font identifier from image works best when the letters are sharp, roughly horizontal and at least 40–50 pixels tall. Crop away busy backgrounds; one or two words with distinctive letters beat a whole paragraph of blurry ones.
2. Let the detector read the letterforms. Upload the picture above. The detector finds the text in the image, isolates each letter’s shape and works out which visual family it belongs to — serif, sans-serif, script, display — before scoring the closest candidates from real fonts.
3. Compare candidates side by side. Type the same words you saw and preview them in every candidate, rendered live with the actual font files. Putting your sample next to a match is the quickest way to check a font image against reality — differences in a single letter jump out immediately.
4. Leave with a name — or with the font itself. If a candidate matches, you have your answer and can go license or download it. If nothing matches, you are not stuck: upload the same picture to Font from Image and YoFont recreates the style as a complete, installable font.

The whole flow is free and needs no account. And unlike a plain reverse-lookup, it does not dead-end when the honest answer is “this typeface is not in any catalog” — which, as you will see below, happens far more often than people expect.

Reading letterforms like a font detector

Whether the work is done by software or by your own eyes, identifying a font comes down to the same skill: knowing which details actually separate typefaces. Thousands of fonts look “about the same” at arm’s length; almost none look the same in these seven places. This is the checklist a font detector effectively runs through — and the one to use yourself when a photo is too rough for automation.

• The lowercase a and g. The single most diagnostic pair in Latin type. A double-story a (with an upper bowl and a lower bowl) versus a single-story a (a simple circle with a stem) instantly splits the candidates in half. The g is even richer: double-story g’s have an “ear” and a lower loop whose shape — open, closed, angular — is close to a fingerprint. When you identify the letter type of a and g first, you rule out most of the library before looking at anything else.
• The tail of the Q. Designers treat the Q as their signature. The tail can cross the bowl, sit entirely outside it, curl back like a comma or sweep under the following letter. Two fonts identical everywhere else routinely give themselves away here.
• Vertices of M, W and N. Does the middle of the M reach the baseline or stop halfway? Are the outer strokes vertical or splayed? Are the joins sharp points, flat cuts or blunt overlaps? These apex details survive even low-resolution photos.
• Serifs — and their brackets. Not just “has serifs / doesn’t”. Slab serifs are heavy rectangles; hairline serifs are thin dashes; bracketed serifs flow into the stem with a curve, unbracketed ones meet it at an angle. Wedge-shaped serifs point to entirely different families than flat ones.
• Stroke contrast and its axis. Compare the thickest and thinnest parts of an O. High contrast with a vertical axis says modern didone; gentle contrast on a tilted axis says old-style; almost no contrast says geometric or grotesque sans. Contrast is visible even when serifs are too small to judge.
• Apertures of c, e, s and a. How far open are the “mouths” of these letters? Wide-open apertures feel humanist and friendly; nearly closed ones feel neutral and engineered. The angle of the e’s crossbar — level or tilted — is a bonus clue from the same group.
• Terminals, tails and the odd ones out. The leg of the R (straight, curved, kicked out), the top of the t and f, ball terminals on the r and y, the shape of the ampersand, and the numerals — old-style figures that dip below the baseline versus lining figures that don’t. When two candidates still tie, one of these breaks it.

This list is also the honest answer to every “name that typeface” quiz: professionals don’t recognize fonts whole, they read four or five of these markers and eliminate. With the checklist in hand you can narrow thousands of candidates to a shortlist before any software gets involved — and judge the software’s suggestions instead of trusting them blindly.

Identify a font from an image in any script

Here is the quiet limitation of almost every tool in this category: they are built for Latin letters. Feed a typical font identifier Cyrillic, Arabic or Chinese and it either refuses, or confidently “matches” your Ж against a K. The letter-anatomy checklist above is Latin-shaped too — Arabic has no lowercase, Chinese has no Q, Devanagari hangs from a headline instead of sitting on a baseline. Different scripts need different eyes.

YoFont was built multilingual from day one, so the identify-font-from-image workflow here is not Latin-only. Upload a sample in Cyrillic, Greek, Arabic, Hebrew, Devanagari, Chinese, Japanese or Korean and the same input–output contract holds: picture in, style analysis out — and, when you want it, a working font in that style back. Nine writing systems are supported end to end, and a recreated family can cover up to 23 languages at once.

This matters even for purely Latin projects. Suppose you do name the typeface from a photo — then discover it ships with no Cyrillic, patchy Greek and no Arabic at all, while your product launches in six markets. Identification answers “what is it?”; only recreation answers “can I actually use this style everywhere I need it?”

When the font can’t be named

A surprisingly large share of the type people photograph has no name to find. Logo lettering is usually drawn by hand for one client and never released. Studios routinely license a font and then modify it — narrower, rounder, custom letters — so the original no longer quite matches. Signage may predate digital type entirely, and plenty of fonts from defunct foundries or one-off student projects simply never made it into any searchable catalog. Every classic identifier hits the same wall here: no name exists, so no lookup can return it.

This is where this page stops being a dead end. Upload the same picture to Font from Image and instead of a name you get the thing you actually wanted: a new, original font in that style. Not a picture of letters — a real font family with a full character set, kerning and ligatures, covering up to 23 languages, exported as TTF, OTF, WOFF and WOFF2 so it installs on desktop and embeds on the web.

Often that beats finding the name. A successful identification hands you someone else’s font — possibly expensive, possibly licensed per-seat or per-pageview, possibly missing half the languages you need. A recreated style is yours: consistent weights, editable glyphs, one license you control. For a brand asset, “inspired by what I saw, built for what I ship” is usually the stronger position.

Free font finder vs paid tools: three ways to get an answer

Everything on the market boils down to three approaches. All three are legitimate; they just answer different questions, cost different amounts and fail in different places.

A practical rule of thumb: if the sample looks like a mainstream retail font, the detector above will likely name it in seconds. If it looks custom, obscure or non-Latin, skip the frustration — recreate the style and get working files the same day. And if you are simply curious what it is, the forums are still a lovely corner of the internet.

相關工具