Languages GotThat supports, what “auto-detect” means, and where it is weakest
One honest page instead of a hundred “Japanese to English live translation” pages. What you can read subtitles in, what the people around you can speak, how fast it is, and the situations where it gets things wrong.
GotThat shows subtitles in five languages: Chinese (Simplified), English, Japanese, Thai and Turkish. You choose one of those to read. The language being spoken is detected automatically per sentence, so a table mixing Japanese and English, or Turkish and English, works without any setting. Subtitles appear about two seconds after speech. It needs an internet connection.
Subtitle languages (what you read)
| Language | Code | Notes |
|---|---|---|
| English | en | Default on the English site |
| 中文(简体) | zh-Hans | Simplified Chinese; Traditional is not a separate option yet |
| 日本語 | ja | Reply suggestions and recap also in Japanese |
| ไทย | th | Thai interface and subtitles; not yet reviewed by a native speaker |
| Türkçe | tr | Turkish interface and subtitles; not yet reviewed by a native speaker |
Spoken languages (what the people around you say)
The same five are the languages we test and tune for, and the ones the three helpers (Topic, Missed, Reply) are designed around. The speech recognition model behind GotThat recognises more languages than that, so other languages will often produce subtitles, but we do not promise it and we do not list them until we have tested them at a real table. The language is detected sentence by sentence; you never pick “Japanese → English” in advance.
Speed
Under normal conditions the first words of a sentence appear about two seconds after the person starts speaking, and the sentence is finalised shortly after they stop. Fast enough to follow and to know a question is for you; not fast enough to interrupt someone mid-sentence.
Where it is weakest (read this before you rely on it)
- Many people at once. Speaker labels are most reliable with two or three people taking turns. With five people, overlapping talk, or two people with very similar voices, lines will sometimes get the wrong label. Tap the name on that line to correct it.
- Loud rooms. A busy izakaya or a street-side table degrades recognition. Move the phone closer to the speakers; it helps more than you would think.
- Names, jargon, in-jokes. Proper nouns and company-internal words are often transcribed phonetically. You can tap a line to see the original.
- No offline mode. Nothing runs without internet.
- Not a transcript for the record. It is built to tell you what is being discussed and whether you are being asked something, not to be quoted in a dispute.
Privacy, briefly
Audio is streamed to our servers for recognition and translation and is not kept unless you tap “Save” at the end of a session. We do not build voiceprints of the people you talk to. Details in the privacy policy.
10 free minutes a day, no sign-up. Pick the language you read and press start.
Questions people ask
Can two people read subtitles in different languages?
Each phone shows one language. Two people who need different languages each run it on their own phone.
Will you add more subtitle languages?
Yes, when we can test them at a real table and have the interface reviewed by a native speaker. We would rather list five that work than fifty that might.
Does it translate text or photos?
No. GotThat is for live speech. For menus and signs, use a camera translator.