HTML Entity Encoder/Decoder

Choose one literal-text conversion

1. Paste text and choose Encode or Decode. The same input stays when switching direction; output is never silently moved into input. Empty input cannot convert, but spaces, tabs and newlines are meaningful data. Input CRLF and lone CR become LF.
2. For Encode, Keep Unicode escapes six punctuation characters and keeps other admitted Unicode. ASCII references also encodes non-ASCII characters as references, with hexadecimal, decimal or named-preferred output and hexadecimal fallback. For Decode, choose Text segment or Attribute-value text; this controls reference ambiguity, not HTML document parsing.
3. Select Encode text or Decode references. Read the literal result and its UTF-8 byte and Unicode-scalar counts. Copy and Download use the complete current result even when the ordinary preview is shortened. Decode runs once, then normalizes any newly decoded CR or CRLF to LF.

Understand references without confusing contexts

Keep Unicode replaces ampersand, less-than, greater-than, double quote, apostrophe and backtick. Accents, emoji and other admitted Unicode can remain literal in UTF-8 text. ASCII references is a representation choice, not a requirement to encode every letter or a promise of safety in every HTML, URL, JavaScript or CSS context.

Named references are case-sensitive and can produce more than one Unicode scalar. Legacy missing-semicolon references may be recovered differently in text and attribute-value contexts. Unknown or incomplete references can stay literal. This pinned recovery behavior does not validate a document or report every current tokenizer error.

Decode treats the whole argument as reference-bearing text. Tags and quotes are not parsed, and even script-looking text stays inert output here. Decoding &amp;lt; once leaves &lt; as literal text instead of decoding it again to <. Encoding already-written references escapes their ampersands again because they are input data.

Both input and result use LF line endings. Decode can turn a numeric CR reference into CR internally, but this tool then changes it to LF; a decoded CRLF pair becomes one LF. There is no other trimming, Unicode normalization, signature removal or added final newline. The result is not a sanitizer or an HTML validity verdict.

Follow source, conversion and output

The Encode example is <b>café & 😀</b>, 19 UTF-8 bytes. Keep Unicode produces &lt;b&gt;café &amp; 😀&lt;/b&gt;, 35 bytes and 31 Unicode scalars. Its 32 UTF-16 units differ because the emoji occupies a pair. Markup remains text rather than becoming a live element.

For that same source, ASCII references produces 53 bytes in Hexadecimal, 48 in Decimal and 46 in Named preferred. Named preferred can use a named accent reference while falling back to hexadecimal for the emoji. These are textual representations, not stronger security modes or claims that one spelling is the only valid reference.

The Decode example is 39 ASCII bytes and recovers a literal b-tagged line followed by CRLF, 15 bytes. One CRLF normalization makes the current result 14 UTF-8 bytes with LF. Text decoding of &amp=1 gives &=1, while Attribute-value text retains the original sequence; neither mode locates actual attributes in the source.

Explicit Unicode, size and runtime bounds

Canonical input fits 100000 UTF-16 units and 262144 UTF-8 bytes. A raw edit candidate above 200000 units, or still above 100000 after LF conversion, is rejected visibly without silently cutting it. Semantic or byte-limit errors leave the bounded draft editable. These units are not words or user-perceived characters.

Source and output reject isolated surrogates, C0 controls except tab and LF after line-ending normalization, DEL and C1 controls, U+FDD0 through U+FDEF and scalars ending in FFFE or FFFF. Other admitted combining, RTL, joining and formatting characters remain data. This is a declared narrower text profile, not a claim that every rejected character is invalid HTML or Unicode.

Recovery can itself create a forbidden output character. The reference for 128 becomes the euro sign, but references producing U+0081, U+0001 or U+FDD0 fail output validation. Zero, surrogate-valued and out-of-range numeric references recover U+FFFD and can be accepted. No characters are silently removed to force an output through this profile.

The complete result fits 1048576 UTF-8 bytes and UTF-16 units. The ordinary preview shows at most 100000 Unicode scalars and marks any clipping; full Copy, its readonly fallback and Download retain the complete current result. A 15000 millisecond worker watchdog produces no partial export on failure. Library intermediates can be larger than the final result cap; these bounds are not memory or throughput certification.

Local conversion and current text

This tool adds no upload, account, saved history, input logging or text-bearing share URL. Conversion runs in a bounded browser worker; edits, options, cancellation and clearing retire obsolete results and export feedback. Existing site services remain subject to site policies; local processing does not certify those services or deployment.