Encoding Auto-Fix

An Obsidian plugin that automatically converts non-UTF-8 text files to clean UTF-8 the moment they are added to your vault — before Obsidian or a sync plugin can corrupt the content.

If you have ever dragged a .txt into Obsidian and seen your Korean/Japanese/Chinese text turn into garbage like ��� or 뿯붿, or seen Self-hosted LiveSync refuse a file with File seems to be corrupted! Writing prevented, this plugin fixes the root cause.

---

The problem (does this sound familiar?)

Obsidian reads and writes every note as UTF-8 and does not support other encodings. So when a text file uses a different encoding, the bytes are mis-decoded the instant Obsidian touches them:

• A UTF-16 file (e.g. a transcript, or a .txt saved as "Unicode" from Windows Notepad) gets its multibyte characters destroyed — replaced with the Unicode replacement character � (U+FFFD).
• A EUC-KR / CP949 file (the legacy Korean encoding, still produced by lots of Korean software) shows up as mojibake.
• With a sync plugin like Self-hosted LiveSync, the damage gets worse: the byte count no longer matches, you see errors like seems to be corrupted! Writing prevented (59286 != 100678), and the broken/blocked file propagates to your other devices.

Once � has been written to disk, the original characters are gone forever — different source bytes all collapse to the same �, so there is nothing left to recover. The only real fix is to catch the file before that happens. That is exactly what this plugin does.

Symptoms this addresses

• Korean / CJK text shows as �, ?, or mojibake after import or sync
• File seems to be corrupted! Writing prevented (X != Y) in Self-hosted LiveSync
• A file's byte size mismatches between devices (recordedSize vs storageSize)
• .txt exported from transcription tools, call recorders, Notepad ("Unicode"/"Unicode big-endian"), or old Korean apps looks broken in Obsidian

---

How it works

1. Listens for the vault create event (a file being added/imported).
2. Reads the raw bytes with vault.readBinary() — this avoids Obsidian's lossy UTF-8 text read, so the original bytes are still intact.
3. Detects the real encoding:
   • BOM check → UTF-8 BOM, UTF-16 LE, UTF-16 BE
   • Null-byte distribution heuristic → BOM-less UTF-16 LE/BE
   • UTF-8 validity check → leave clean UTF-8 alone
   • EUC-KR / CP949 / UHC fallback (WHATWG euc-kr decoder, which includes the CP949/UHC extension) with a quality check
4. If the file is not clean UTF-8, it decodes with the correct decoder and rewrites the file as UTF-8 (BOM stripped).

Features

• ✅ Automatic conversion on file creation — just drag the file in.
• ✅ Detects & converts: UTF-16 LE/BE (with or without BOM), UTF-8 with BOM, EUC-KR / CP949 / UHC.
• ✅ Safe by design: files that are already clean UTF-8 are never touched. Encodings it can't confidently identify are left alone with a warning — it never makes a destructive guess.
• ✅ Manual command: "현재 파일을 UTF-8로 변환 (Convert current file to UTF-8)" for fixing a file on demand.
• ✅ Works on desktop and mobile.

Handled text extensions: txt, md, markdown, csv, tsv, json, log, text, srt, vtt.

---

Installation

Community plugins (once approved)

Settings → Community plugins → Browse → search "Encoding Auto-Fix" → Install → Enable.

Via BRAT (beta, available now)

1. Install the BRAT plugin.
2. BRAT → Add beta plugin → enter kathar0s/obsidian-encoding-autofix.

Manual

1. Download main.js and manifest.json from the latest release.
2. Put them in <your-vault>/.obsidian/plugins/encoding-autofix/.
3. Reload Obsidian → Settings → Community plugins → enable Encoding Auto-Fix.

---

Important notes & limitations

• 🔌 The plugin must be installed on the device where the file is added. .obsidian/plugins/ is not always shared by sync, so install it on each device where you import files.
• 🧨 Already-corrupted files cannot be recovered. If � characters were already written to a file, the information is lost — re-import or fix the source file (e.g. re-export it as UTF-8) instead.
• 🇰🇷 EUC-KR / CP949 detection uses a confidence check. In the rare case a file can't be confidently identified, it is left untouched and you get a warning so you can convert it manually (no destructive guessing).
• This plugin only converts file encoding. It does not change file contents otherwise.

---

Why I built this

I lost a meeting transcript to this exact bug: a UTF-16 .txt synced through Self-hosted LiveSync, got blocked with Writing prevented, and when force-received, every Korean character had become �. After confirming the original was fine and only Obsidian's UTF-8 assumption was the problem, I built this so that dragging a file in "just works" — and so others hitting the same wall don't lose their notes. If it helped you, an issue or star is appreciated. 🙏

Contributing / Issues

Bug reports and PRs welcome at the GitHub repo. If a specific file isn't detected correctly, please attach the first ~32 bytes (hex) and the encoding you expected.

License

MIT

---

한국어 설명

옵시디언에 추가되는 비-UTF-8 텍스트 파일을 자동으로 UTF-8로 변환하는 플러그인입니다. 옵시디언이나 동기화 플러그인(예: Self-hosted LiveSync)이 내용을 깨뜨리기 전에 가로채 고칩니다.

이런 증상이 있다면

• .txt를 옵시디언에 넣었더니 한글이 � 또는 뿯붿 같은 깨진 문자로 보임
• Self-hosted LiveSync에서 File seems to be corrupted! Writing prevented (59286 != 100678) 오류로 파일이 안 받아짐
• 기기마다 파일 크기가 다르게 잡힘 (recordedSize ≠ storageSize)
• 녹취/통화 녹음 프로그램, Windows 메모장("유니코드"로 저장), 오래된 한국 소프트웨어가 만든 파일이 옵시디언에서 깨짐

원인

옵시디언은 모든 노트를 UTF-8로 가정하고, 다른 인코딩을 지원하지 않습니다. 그래서 UTF-16이나 EUC-KR/CP949 파일은 옵시디언이 읽는 순간 한글이 �(U+FFFD)로 손상됩니다. 한 번 �로 바뀌면 원래 글자는 영구 소실됩니다(서로 다른 바이트가 모두 같은 �로 뭉개지기 때문). 그래서 손상되기 전에 잡아야 하고, 이 플러그인이 그 역할을 합니다.

동작 방식

1. vault create 이벤트 감지
2. readBinary()로 raw 바이트를 먼저 읽음 (옵시디언의 손실성 UTF-8 읽기 회피)
3. BOM → null 바이트 분포 → UTF-8 유효성 → EUC-KR/CP949 순으로 인코딩 감지
4. 비-UTF-8이면 올바른 디코더로 읽어 UTF-8(BOM 제거)로 재저장

지원 인코딩

• UTF-16 LE / BE (BOM 유무 무관)
• UTF-8 BOM (BOM 제거)
• EUC-KR / CP949 / UHC (확장 완성형 포함)

안전 장치

• 이미 깨끗한 UTF-8 파일은 건드리지 않음
• 확실히 판별되지 않는 인코딩은 변환하지 않고 경고만 표시 (잘못된 추측으로 파일을 망치지 않음)
• 수동 명령: 현재 파일을 UTF-8로 변환

주의

• 파일을 추가하는 기기마다 플러그인을 설치해야 합니다 (.obsidian/plugins/는 동기화가 안 될 수 있음).
• 이미 �로 깨진 파일은 복구 불가 — 원본을 UTF-8로 다시 내보내 재추가하세요.