People often say a file "needs conversion" when the real problem is encoding or fonts. The symptoms look similar if you do not read Chinese every day. The fixes do not. Running OpenCC on a broken file will not recover the author's characters. It will map nonsense to different nonsense.
Work through the symptoms below before you upload anything.
Boxes (tofu)
A hollow box or tofu square usually means the character exists as a code point but the current font has no glyph. The data may already be correct. Change the font to one with CJK coverage. See Chinese fonts and Unicode.
If every Chinese character is a box, the font is the first suspect. If only some characters are boxes, you may have a mix of rare graphs, a font subset, or a regional glyph the font never included.
Do not convert yet. If the boxes hide Traditional characters you already wanted, conversion is the wrong lever.
Question marks
A ? that replaced a character often means the text was forced through a charset that could not store it. Saving a UTF-8 Chinese sentence as Latin-1, or pasting through an old SMS gateway, is a typical path. The original character may already be gone. Conversion cannot reconstruct it.
If you still have the source email or the original .docx, go back to that. The copy with question marks is a corpse.
Mojibake
Strings such as 涓浗 instead of 中国 are UTF-8 bytes read as the wrong encoding, or the reverse. The page is full of rare or nonsense characters that a Chinese reader immediately rejects.
Re-open the file with UTF-8 if you still have the source. In Word, try the encoding dialog when opening a plain-text file. In a browser, try the encoding menu if you are looking at an old site.
If only the broken copy exists, automatic repair is hit-or-miss. Some tools guess UTF-8 versus GBK versus Big5. Guessing can make things worse. Work on a copy.
Mojibake is listed in common conversion problems because people convert it by accident and then have two problems.
Mixed scripts that look "almost Chinese"
Japanese sentences can look related. They are not safe OpenCC input. If a file is bilingual, extract the Chinese part. Korean hanja is the same warning.
A PDF of a photograph is not garbled in the encoding sense. It is not text. See the PDF guide.
When conversion is the right next step
If you can read the Chinese, and you only want Simplified or Traditional forms, use the document converter or text converter. If you cannot read it, fix encoding or fonts first.
A short test: copy one sentence into the text converter. If the input box already shows garbage, stop. If the input is readable and the output is the other standard, the file was fine and you can convert the rest.
A diagnosis order
- Can you highlight and copy a real Chinese sentence? If no, you have a scan, outlines, or a font so broken you cannot see the text.
- Does changing the font make the sentence appear? If yes, you had boxes, not a conversion job.
- Does the copy look like random rare characters? If yes, you have mojibake.
- Are there question marks in the middle of words? If yes, data was already lost.
- Only then pick an OpenCC type and convert.
Frequently asked questions
My colleague sees Chinese and I see boxes. Who is right? Both. The file can be correct and your font coverage can be wrong. Install a CJK font before you ask them to convert again.
Can this site repair mojibake? No. It converts readable Chinese standards.
Why did a web page look fine and the downloaded file look wrong? The browser guessed an encoding. The saved file froze the wrong guess.
Is Big5 the same as Traditional Chinese? Big5 is an old encoding used for a lot of Traditional text. Unicode Traditional characters are the modern storage. A Big5 file mis-opened as UTF-8 becomes mojibake.
Should I convert a file to "fix" question marks? No. Recover the source.