Garbled contact names or an invalid UTF-8 byte error mean the .vcf file was written in one character encoding and read as another. The pattern of the garbage tells you which encoding is wrong. Jump to your situation below or work through the methods in order.
Garbled names in a vCard come down to one thing: a character encoding mismatch. The file was saved in encoding A but the importing app read it as encoding B, so the bytes map to the wrong characters. The pattern tells you what went wrong. The classic A-tilde pattern (Ã, Ã…) is a Windows-1252 file read as UTF-8. Question marks or empty boxes mean a CJK encoding (Chinese, Japanese, Korean) read as a Western one. Raw sequences like =C3=BC are undecoded quoted-printable from an old vCard 2.1 file. Odd unrelated symbols can be MacRoman read as UTF-8. The fix is always the same principle: identify the real encoding, then convert the file to UTF-8. Remove any per-property CHARSET and ENCODING=QUOTED-PRINTABLE parameters, upgrade vCard 2.1 to 3.0 which uses plain UTF-8, and save without a byte order mark for the widest compatibility.
Every text file stores characters as bytes, and an encoding is the rulebook that maps bytes to characters. When the app that wrote the .vcf and the app that reads it agree on the rulebook, names display correctly. When they disagree, you get garbage, or the reader hits a byte that is not valid in the encoding it assumed and reports an invalid UTF-8 sequence.
vCard makes this worse because the format changed over versions. vCard 2.1 does not declare a file-level encoding and often uses quoted-printable, while 3.0 and 4.0 use plain UTF-8. So an old 2.1 file moved between systems is the classic source of garbled names, and the symptom pattern points straight at the encoding to fix.
“The garbage is a fingerprint, once you read it you know the fix. A capital A with a tilde in front of every accented letter is always Windows-1252 being read as UTF-8. Question marks mean a Chinese, Japanese or Korean file got read as Western. And =C3=BC sitting in a name is quoted-printable that nobody decoded, which screams old vCard 2.1. So I do not guess, I match the pattern, convert the whole file to UTF-8, and upgrade it to 3.0 so the per-property encoding nonsense goes away for good.”
file --mime-encoding yourfile.vcf
iconv -f WINDOWS-1252 -t UTF-8 input.vcf -o output.vcf
=C3=BC style sequences, the file uses quoted-printable that was not decoded.ENCODING=QUOTED-PRINTABLE and CHARSET parameters from each line.VERSION:2.1 with VERSION:3.0.The symptom is the diagnosis: Ã means Windows-1252 read as UTF-8, question marks mean a CJK encoding, and raw =C3=BC means undecoded quoted-printable. Whatever the source, the cure is to convert the file to UTF-8 and, where possible, upgrade it to vCard 3.0. Always save UTF-8 without a BOM for the widest compatibility.
If names are still wrong after converting, the file may have been double-encoded or built by concatenating files with different encodings, so a single conversion cannot fix it. Re-export the contacts cleanly from the original source as vCard 3.0 in UTF-8, or run them through a converter that detects and normalizes the encoding. You can also submit your error to us for a tailored fix.