Google Translate conversation mode with three or more people: why it struggles, and what to use instead
Google Translate's conversation mode is good at what it was designed for: two people, two languages, taking turns. Put it in the middle of a table of four and it starts to miss who said what. Here is why, and what a phone-on-the-table approach does differently.
Short answer: Google Translate conversation mode works best for a two-person exchange where each person waits for the other. It does not label speakers, so in a group every sentence looks like it came from the same person, and overlapping talk gets merged or dropped. GotThat listens to the whole table, labels speakers A, B and C, and keeps subtitles flowing without anyone taking turns.
What conversation mode actually does
In the Google Translate app, conversation mode shows two languages at once. You tap the microphone (or let it auto-detect) and it listens, transcribes, translates, and can read the translation aloud. It is free, works for a long list of languages, and is excellent at a hotel desk, a pharmacy or a taxi. Google's own help page describes it as a way to “talk with someone” in another language: one person at a time.
What happens at a table of four
- No speaker labels. The transcript is one stream. When A asks a question and B answers, you see two translated lines with nothing to tell you who said which, so you cannot tell whether the question was for you.
- Overlap. Two people talking at once is normal at dinner. Conversation mode is built around one microphone turn at a time; overlapping speech tends to be merged into one sentence or lost.
- The phone becomes the conversation. To keep it accurate you end up tapping, holding the phone up, asking people to repeat. The others start waiting for your phone, which is exactly the thing a group chat cannot survive.
- No catch-up. If you look away for a minute there is no “what did I miss”, only a scrolling log.
Side by side
| Feature | Google Translate conversation mode | GotThat |
|---|---|---|
| Designed for | Two people taking turns | Two to five people talking normally |
| Speaker labels | No | Yes (A, B, C…; rename them) |
| Overlapping speech | Merged or dropped | Handled, with occasional wrong labels in noise |
| What the others do | Wait for your phone, sometimes speak into it | Nothing; the phone just lies on the table |
| Catch-up / topic summary / reply help | No | Three buttons |
| Languages | 100+ (text), dozens for speech | Subtitles in 5: Chinese, English, Japanese, Thai, Turkish; speech detected automatically |
| Offline | Yes, with downloaded packs | No |
| Price | Free | 10 free minutes a day, no sign-up |
| Where it runs | Android, iOS | Browser (any device), iOS; Android coming |
When Google Translate is still the better choice
Use conversation mode when it is you and one other person, when you need a language GotThat does not show subtitles in, or when you have no signal and downloaded an offline pack. It is also the better tool for reading signs and menus with the camera. GotThat is not trying to replace it for those jobs.
What a group session looks like in GotThat
The third line is a question for you. Tap “Reply” and you get a few things you could say, in Japanese, with the English underneath.
Try it at the next dinner. Open the web app, put the phone on the table, press start.
Questions people ask
Can Google Translate conversation mode handle multiple people?
It can transcribe whoever speaks into the phone, but it does not separate speakers and is built around one turn at a time. For a group you will see one merged stream without names.
Does GotThat need the other people to install anything?
No. Only your phone runs it. The others talk as they normally would.
How accurate are the speaker labels?
Most reliable with two or three people taking turns. With many people, a loud room or similar voices it can mislabel a line; tap the name on that line to fix it.
Is audio stored?
Not by default. You get a recap when you stop; only if you tap “Save” are the audio and text kept, in your account.