What makes a good slug
A slug is the readable last part of a web address, such as my-first-blog-post in example.com/blog/my-first-blog-post. A good one is short, in lower case, uses only letters, digits and one kind of separator, and tells a person what the page is about before they open it.
Use hyphens to separate words on the web. Google's guidance on URL structure recommends hyphens over underscores, and hyphens are what most content management systems produce. Underscores and dots suit file names, database keys and code identifiers, where a hyphen could be read as a minus sign.
Choose a slug once and keep it. If a page's address changes after it has been shared or indexed, old links break unless you add a redirect.
How accents and other alphabets are converted
Accented letters are split into the plain letter and the accent mark, and the mark is dropped, so é becomes e. A few letters are not made of a letter plus a mark and have their own rules: ß becomes ss, æ becomes ae, ø becomes o, đ becomes d, ł becomes l, þ becomes th and ı becomes i. German umlauts become a, o and u by default, or ae, oe and ue if you switch that option on.
Other alphabets are converted letter by letter into a simple Latin spelling meant for web addresses. Russian follows the common style (ж is zh, х is kh, ц is ts, щ is shch, and the hard and soft signs are dropped). If the text has Ukrainian letters, г becomes h, и becomes y, and Київ becomes kyiv. Greek follows a simplified form of the ELOT 743 standard, with η as i, υ as y and ου as ou. Japanese kana use Hepburn style (し is shi, ち is chi, つ is tsu) and the long vowel mark is dropped, so ラーメン becomes ramen.
Korean syllables are written with the Revised Romanization letter by letter, so 한국어 becomes hangukeo. Devanagari uses simple rules that drop the final inherent a, so भारत becomes bharat. Arabic, Persian and Hebrew are written without most vowels, so the result is the consonants only.
What this tool cannot do
Chinese characters, Japanese kanji, Thai, Khmer, Tamil and most other scripts need a dictionary to be read aloud correctly, and this tool has none. It does not guess. Those characters are listed above the result, and left out of the slug unless you turn on Keep letters that cannot be converted. In mixed text such as 東京タワー, the kana part (タワー) is converted and the kanji are listed as unsupported.
Korean and Devanagari are converted by rule, not by dictionary, so sound changes between syllables and some medial vowels are not reflected. Names and brand spellings in these scripts are best typed in Latin letters by hand.
The list of common words to remove is English only, and it only removes whole words that match.
Slugs in your own alphabet
Turn off Convert other alphabets to Latin letters to keep letters as typed, and привет мир becomes привет-мир. Modern browsers show such addresses as written, but the address is sent as percent codes such as %D0%BF%D1%80, and some tools, spreadsheets and older systems display or break it. Choose a Latin slug when the link will be copied into many different places, and a native one when your readers expect to see their own language.