URL Slug Generator - Clean Permalinks From Any Alphabet

The same article, two addresses

An address is the one part of a page that outlives the page. Titles get rewritten, images get swapped, the layout gets rebuilt twice, and the path stays, because every link anyone ever made points at it. That is why the twenty seconds spent turning a headline into a path are worth more than they look.

Left to the CMS

/index.php?p=482&cat=7&lang=de

Nothing in it says what the page is about
Two titles can quietly land on the same path
Accented letters get deleted rather than converted
Shared in a chat window, it looks like spam

Written on purpose

/schoene-gruesse-fuer-kaeufer

Readable before anyone clicks it
Collisions are caught while you can still rename
Every letter has a defined Latin form
Survives being pasted into a plain text email

Paste one headline or two hundred, one per line. Each line comes back as a path, with a note saying what changed on the way: which symbol was spelled out, which stop word went, where the length cap fell, and which line collided with a line above it.

What a title has to survive on its way to a path

A path may contain lowercase letters, digits and one separator. Everything else has to become one of those three or disappear, and the interesting question is which. Deleting is the lazy answer and it is the one that loses data: drop the ring from å and Malmö still reads, drop the caron from š and Ceška becomes Ceka, which is a different word.

The order of operations here is deliberate. Letters that carry a mark are decomposed into a base letter plus that mark, and the mark is discarded, so é becomes e, ș becomes s and ế becomes e without needing an entry per letter. Letters that carry no mark at all get an explicit rule, because there is nothing to strip: ß is not an s with something on top, it is its own letter, and it becomes ss. The same goes for ø, ł, đ, þ and æ.

One title, traced through every stage. Starting from Straße des 17. Juni:

Letters with no decomposition are spelled out: Strasse des 17. Juni
Remaining marks are decomposed and dropped: nothing left to do here
Lowercased: strasse des 17. juni
Every run of anything else becomes one separator: strasse-des-17-juni
Length checked against the cap: 19 characters, under 60, left alone

Two smaller decisions do a surprising amount of work. An apostrophe is closed up rather than broken on, so Don’t Panic gives dont-panic and not dont-t-panic. And four symbols are spelled out instead of dropped, because dropping them changes the meaning: The Best 50% Off Deal becomes the-best-50-percent-off-deal, where deleting the sign would have left the-best-50-off-deal.

Six settings, top to bottom

Titles - one per line. A whole editorial calendar can go in at once, up to 200 lines, and duplicates are only visible when the batch is judged together rather than one title at a time.
Separator - hyphen or underscore. Hyphen for anything a search engine will read, underscore where a downstream system treats a path segment as an identifier. Whichever you pick, the other one stops appearing in the output, which is where most hand written slugs go wrong.
Accents - drop the mark, or use the German convention. The first gives schone-grusse, the second gives schoene-gruesse. German readers search for the second spelling, so on a German site the convention is not decoration.
Length cap - a number of characters, or 0 for none. The cut lands on the last whole word that fits, never mid word, unless the first word is already longer than the cap.
Stop words - keep or remove. Removing turns a 76 character path into a 47 character one. It also removes meaning from some titles, which is why it is a switch and not a default.
Duplicate handling - add a numeric suffix, or flag and leave alone. Flagging is the honest setting when you still have the option of changing the title.
Read the table - the slug column is what to paste, and the last column says what happened to get there.

What each writing system turns into

Latin letters with marks are handled by decomposition, so the table below lists only the cases that need a rule of their own. Cyrillic and Greek get a single romanization applied to every language that uses them, which is a deliberate simplification and is worth knowing about before you trust it with a Ukrainian title.

Input	Becomes	Worked example
Any letter with a mark	base letter	Ceška Košarka → ceska-kosarka
ß	ss	Straße des 17. Juni → strasse-des-17-juni
ä ö ü in German mode	ae oe ue	Schöne Grüße → schoene-gruesse
ø ł đ þ æ œ	o l d th ae oe	Zażółć gęślą jaźń → zazolc-gesla-jazn
Cyrillic	one romanization	Москва Россия → moskva-rossiya
Cyrillic digraphs	zh ch sh shch yu ya	Ёлка и Чайка → yolka-i-chayka
Greek	one romanization	Αθήνα Ελλάδα → athina-ellada
& % + @	and percent plus at	C++ Tutorial → c-plus-plus-tutorial
Apostrophes and quotation marks	closed up	Don’t Panic → dont-panic
Punctuation, emoji, currency, degrees	one separator	Summer 🌞 Sale → summer-sale
Everything else	reported, not hidden	A CJK title returns empty and says which characters had no Latin form

What the careless version costs later

None of these are hypothetical. They are the four ways a path goes wrong, and every one of them is cheap to prevent and expensive to undo, because undoing means a redirect that has to be maintained for as long as anyone still links to the old address.

What goes wrong	Left alone	Caught here
Two titles, one path	The second save overwrites the first, or the CMS appends a suffix nobody chose	The collision is named while you can still rename the title
A letter deleted rather than converted	ceska becomes ceka, and the word no longer matches anything anyone searches for	Every mark bearing letter keeps its base letter
Mixed separators in one path	already-hyphenated_title, which is two conventions in one address	The chosen separator is the only one that appears
A path longer than the column	The insert fails, or the value is silently cut mid word by the database	The cut lands on a word boundary and is reported
Cost of fixing it after launch	A permanent redirect per address, kept forever	Nothing, the title was still a draft

Where the cap falls, on one real headline

The same 76 character headline, run at four settings. Nothing here is estimated, every row is what the tool returns for The Complete Guide to Choosing Running Shoes for Flat Feet and Overpronation.

Setting	Result	Chars
No cap	the-complete-guide-to-choosing-running-shoes-for-flat-feet-and-overpronation	76
Cap 60	the-complete-guide-to-choosing-running-shoes-for-flat-feet	58
Cap 40	the-complete-guide-to-choosing-running	38
No cap, stop words removed	complete-guide-choosing-running-shoes-flat-feet	47

Notice what the cap at 60 threw away: and overpronation, which is the part of the headline a reader with flat feet was searching for. Removing stop words instead keeps overpronation out of the path but saves 29 characters and holds every noun. Neither is automatically right. The point is seeing the trade before the address is permanent, not after.

A batch with a collision in it. Four product titles that a human would call four different pages:

Red Running Shoes! → red-running-shoes
Red Running Shoes? → red-running-shoes-2
Red Running Shoes → red-running-shoes-3
Blue Running Shoes → blue-running-shoes

Three of the four titles differ only in punctuation, so three of them want the same address. The suffix keeps all three reachable, and the report says plainly that a suffix is a patch and a different title is the actual fix.

The awkward cases, answered

Hyphen or underscore, if I only care about search?

Hyphen. A hyphen is read as a word break, an underscore is not, so red_running_shoes can be read as one long token rather than three words. Underscore earns its place where a path segment is really an identifier being handed to another system, and there the readability argument does not apply anyway.

Why does Kyiv come out as kiyiv?

Because one romanization is applied to every language that uses Cyrillic, and it reads the vowel и the Russian way, as i. Ukrainian romanizes that same letter as y, which is where the official spelling Kyiv comes from. The tool does not detect the language of a title, so it cannot switch conventions on its own. For Ukrainian titles, type the romanization you want and let the tool handle the punctuation and the casing.

Should I remove stop words?

Usually not, and definitely not blindly. Removal is worth it when the cap is tight and the words carry nothing, which is where it saved 29 characters in the table above. It is a bad idea when the small word is the meaning: The Who, To Kill a Mockingbird and anything where on and off are opposites. A title made only of stop words is left intact rather than emptied, which is the one case the tool decides for you.

Is there a correct length for a slug?

There is no published limit that makes one length correct. There are two practical ceilings: the width of the database column the value has to fit in, and the point at which a path stops being readable when it is pasted somewhere. A cap between 50 and 75 characters keeps most headlines whole and keeps the address quotable. Set 0 while drafting to see the full path first, then decide where to cut.

Can I change a slug after the page is live?

You can, and it costs you a permanent redirect that has to survive every future migration. Anyone who linked to the old address is relying on it. That is the whole argument for spending twenty seconds on the path before publishing rather than twenty minutes on redirect rules afterwards.

Why did my title come back empty?

Because nothing in it maps to a Latin letter or a digit. A title written entirely in Chinese, Japanese, Korean, Arabic or Hebrew has no transliteration here, so the honest answer is an empty result and a line naming the characters involved, rather than a silent blank. Those languages need a romanization chosen by someone who reads them, or a path built from a translated title.

Does a number at the start of the path cause problems?

Not for the address itself. 5-best-calculators is a perfectly valid path. It is worth a second look only if your routing treats a numeric first segment as a record id, because then a path that is nothing but digits can be swallowed by a route meant for something else. The tool leaves digits alone and shows you the result, so the check takes a glance.

How many titles can go in at once?

Up to 200 lines in one run, and the duplicate check looks at the whole batch, not at adjacent lines. That is the reason to paste a whole content calendar in one go rather than one title at a time: a collision between line 4 and line 180 is invisible any other way.

Related tools

CSS Gradient Generator

The background behind the headline whose path you just wrote, with the contrast checked at both ends - Open the tool

Random Name Generator

People to populate the pages while the real content is still being written - Open the tool

SEO Meta Tag Generator

The title and description that sit above this address, measured in pixel width rather than characters - Open the tool

UTM Link Generator

Campaign parameters bolted onto the finished address, with the channel named before you publish - Open the tool

Business Name Generator

Names that have to work as a brand and as a path at the same time, checked for both - Open the tool

Table Generator

Turns a pasted range into Markdown, HTML and two other shapes for the article behind the address - Open the tool

QR Code Generator

The same address as a printable symbol, where a short readable path really does mean a smaller code - Open the tool