Привет UTF-8 bytes were decoded as Windows-1251. Return to the original bytes and decode them as UTF-8; do not blindly re-encode the already damaged string.
Reference library
Unicode and encodings: UTF-8, UTF-16, and code points
Code points, UTF-8 bytes, UTF-16 code units, normalization, BOM, and encoding error diagnostics.
From text to bytes
Four different levels
- GraphemeWhat the reader perceives: A, é, or 👩💻.
- Code pointAn abstract Unicode number such as U+1F600.
- Code unitA storage unit: 8 bits in UTF-8, 16 bits in UTF-16.
- ByteThe actual values written to a file or sent over a network.
Encoding table
The same text at every level
Copy a character, its code points, UTF-8 bytes, or UTF-16 code units. Spaces separate individual values.
| Character | Code point | UTF‑8 | UTF‑16 | What it shows |
|---|---|---|---|---|
| Single code pointsA single Unicode character can occupy one to four bytes in UTF-8. | ||||
| Latin AASCII is preserved by UTF-8 and occupies one byte. | ||||
| Cyrillic YaMost Cyrillic letters are encoded with two UTF-8 bytes. | ||||
| Precomposed éThe precomposed accented letter is represented by one code point. | ||||
| Euro signThis Basic Multilingual Plane character occupies three UTF-8 bytes. | ||||
| Han character 中A typical CJK ideograph also occupies three UTF-8 bytes. | ||||
| EmojiA code point outside the BMP: four UTF-8 bytes and a UTF-16 surrogate pair. | ||||
| SequencesOne visible grapheme can consist of several code points. | ||||
| Decomposed éThe letter e plus a combining acute looks like é but remains two code points. | ||||
| Woman technologistTwo emoji are joined with ZERO WIDTH JOINER U+200D. | ||||
| Flag of RussiaThe flag is built from two regional indicators rather than one character. | ||||
| Emoji with modifierA base emoji and a skin-tone modifier form one visible grapheme. | ||||
Practical recipes
Declarations, conversion, and normalization
Copy complete declarations and API calls. Each group solves a different layer of the encoding task.
| Task | Code or bytes | When to use |
|---|---|---|
| Web and HTMLHow to declare UTF-8 and reference a character without typing it directly. | ||
| HTML encoding declaration | Place it near the start of head so the browser decodes the document as UTF-8 immediately. | |
| HTTP header | The server-side response declaration; it must match the actual bytes. | |
| Numeric HTML reference | References a code point in HTML. This is markup syntax, not a separate encoding. | |
| JavaScriptConvert strings to bytes and create characters from code points. | ||
| String → UTF-8 | Returns a Uint8Array containing UTF-8 bytes. | |
| UTF-8 → string | Decodes bytes; fatal: true turns an invalid sequence into an error. | |
| Code point → character | Creates 😀 without manually writing a UTF-16 surrogate pair. | |
| NormalizationTurn equivalent sequences into a predictable form. | ||
| NFC: composed form | A common choice for storage, comparison, and interchange without folding compatibility distinctions. | |
| NFD: decomposed form | Decomposes precomposed characters into a base and combining marks. | |
| NFKC: compatibility | Folds compatibility variants such as fullwidth forms; use only when losing those distinctions is acceptable. | |
| BOM and byte orderSignatures at the start of a stream; UTF-8 usually does not need a BOM. | ||
| UTF-8 BOM | A UTF-8 signature. UTF-8 has no byte order, so add it only when a format or application requires it. | |
| UTF-16LE BOM | Indicates little-endian byte order for UTF-16. | |
| UTF-16BE BOM | Indicates big-endian byte order for UTF-16. | |
Mojibake decoder
What the broken text is telling you
The visible pattern often points to the failed boundary, but recovery depends on preserving the original bytes.
Привет UTF-8 bytes were decoded as Windows-1252 or a similar single-byte encoding. Check charset in HTTP, HTML, and the file import settings.
������ The decoder encountered invalid bytes and inserted U+FFFD REPLACEMENT CHARACTER. Find the original file or stream: once bytes become �, the lost values usually cannot be recovered from the string.
?????? The text was saved through an encoding that cannot represent the required characters. Switch to UTF-8 before writing. If characters are already question marks, recover them from the source or a backup.
Quick notes
Keep text intact across boundaries
- Use UTF-8 by default
- Choose UTF-8 for new files, APIs, databases, and web pages, and declare it consistently at every boundary.
- Keep the original bytes
- When text is corrupted, diagnose the byte sequence before saving or re-encoding the visible mojibake.
- Normalize deliberately
- NFC is a practical default for comparison; NFKC changes compatibility distinctions and is not a universal cleanup pass.
- Count user-perceived characters
- String length, code-point count, and grapheme count can differ for emoji and combining sequences.
Encoding guidance checked against the WHATWG Encoding Standard; normalization and BOM behavior follow Unicode Normalization Forms and the Unicode UTF & BOM FAQ.
Related tools
Continue working with text
Copy a symbol, convert markup, clean a fragment, or keep the result in the notebook.
- Encode HTML entities Turn special characters into safe HTML entities or decode them back.
- Convert Markdown to HTML Prepare HTML markup from a Markdown draft.
- Remove HTML tags Keep readable text and discard markup.
- Change text case Convert letters to upper or lower case.
- Open the online notebook Use copied symbols in a note stored locally in your browser.
FAQ
Questions about Unicode and encodings
What is the difference between Unicode and UTF-8?
Unicode assigns characters abstract code points. UTF-8 is one way to encode those code points as bytes. UTF-16 and UTF-32 are other encoding forms.
Why is JavaScript "😀".length equal to 2?
JavaScript strings expose UTF-16 code units. U+1F600 is outside the Basic Multilingual Plane and therefore uses a surrogate pair: two code units.
Does a UTF-8 file need a BOM?
Usually no: UTF-8 has no byte-order ambiguity. A BOM can act as a signature, but add it only when a consuming format or application requires it.
Can mojibake always be repaired?
Only when the original bytes or a reversible sequence of wrong conversions is available. Replacement characters or question marks often mean information has already been discarded.
When should text be normalized?
Normalize at a deliberate system boundary, commonly to NFC before storage or comparison. Do not repeatedly apply compatibility normalization to arbitrary content.