Eddie AI/ helpOpen Eddie

🚀 Getting started · Last updated July 17, 2026

Languages: transcription and captions

Eddie transcribes and captions 100+ languages — including Chinese, Cantonese, Japanese, Korean and Thai — rendering each in its own script.

100+ languages — transcribed, edited and captioned in-scriptEnglish中文 / 粵語日本語한국어ไทยEspañolFrançaisDeutschPortuguês+ dozens morecaptions wrap and render correctly in each script (Chinese, Japanese, Korean, Thai …)RTL note: Arabic & Hebrew transcribe fine; MP4 caption burn-in for them is on the way

Eddie isn't English-only. Transcription, text-based editing and captions work across a wide set of languages, each in its own writing system.

What's supported

  • 100+ languages. The full tier-1 set includes Chinese (both Mandarin and Cantonese), Japanese, Korean and Thai, alongside the large second tier and regional variants such as Canadian French, Swiss German, Latin-American Spanish, European Portuguese and Swedish.
  • In-script everywhere. Whatever the language, the transcript, search and captions work in that script — you edit by the words as written.

Picking the language

Eddie detects the spoken language on import; if you know it up front you can choose it in the language picker so transcription starts from the right model.

Captions render correctly

Captions for Chinese, Japanese, Korean and Thai wrap at the correct character boundaries using bundled Noto fonts — no spaces forced between characters, no missing-glyph boxes — and karaoke highlighting follows the script. Titles and graphics pick a matching script font from their text too.

Right-to-left (Arabic, Hebrew): today's limit

Arabic and Hebrew transcribe normally and appear in the transcript and editor. What Eddie won't do yet is burn right-to-left captions into an MP4 — the karaoke overdraw misaligns, so it declines rather than ship them crooked. To finish RTL captions, export a timeline to your NLE. Full RTL burn-in is on the roadmap.

Get names right in any language

Attach a Backgrounder before importing so people's names and product terms are spelled correctly regardless of language — see Transcripts and speakers.