Now covering 45 languages, live
Whatever you couldn't follow —now it has bilingual captions.
Capture microphone or desktop audio, read the original transcript beside live translation, and replay local recordings anytime — for meetings, talks, interviews, and videos.
90 minutes free trial
Source
Target
Replay the local recording, click any caption to seek00:03 / 00:28
Recording saved on this device · Click any line to jump to that moment
Source and translation, side by side
The language being spoken on the left, the one you picked on the right.
Microphone and desktop audio capture
For desktop capture, choose a tab or window that is actually playing sound.
Project history, export, and resume
Recordings stay in this browser unless you upload them yourself.
Why choose us
Translated by a large language model, with no host or company to switch it on. Open your browser and what you hear becomes bilingual captions.
| FluentSub | Zoom | Teams | Feishu | |
|---|---|---|---|---|
| Core advantageTranslation | Large-language-model translation: carries on from what came before, translates whole sentences by meaning, and a glossary pins technical terms | Not disclosed whether an LLM is used | Not disclosed whether an LLM is used | Not disclosed whether an LLM is used |
| Core advantageDoes someone have to switch it on? | No one. You turn it on yourself | The host's account needs a translated-captions license and the setting turned on; otherwise it isn't available | Translated captions need the organizer to hold a Premium or Copilot license; otherwise they aren't available | An admin has to enable captions; a transcript afterwards needs the host to record |
| Getting started | Opens in your browser; nothing to install | Only inside a Zoom meeting | Only inside a Teams meeting | Needs the Feishu app and an account |
| What it can listen to | In-person classes, online classes, videos, meetings on any platform | Zoom meetings only | Teams meetings only | Feishu meetings; its recorder can capture in-person audio |
| Languages | 45 languages | About 37 | About 30 | Recognizes speech in Chinese, English and Japanese only |
| Original and translation together | Side by side by default (original left, translation right), plus three other layouts | Optional; two lines stacked at the bottom of the video | The caption area shows one language at a time | Optional; tick a box to show both |
| Afterwards | Every bilingual caption is kept: search it, click a line to replay it, generate notes | Keeps the original-language transcript; translations aren't saved | Keeps the original-language transcript; translations aren't saved | Kept only if the meeting was recorded; the transcript can be translated afterwards |
Comparison as of October 2026, based on each company's website and help center. Zoom, Microsoft Teams and Feishu are trademarks of their respective owners; FluentSub is not affiliated with them.
Two switches for captions that are hard to read
AI polish
Rewrites each finished translation with the surrounding context so it reads smoothly. Meaning and sentence breaks stay as they were; it adds a little delay.
Raw
So, uh, the number we looked at last week, it is tied to churn, and each group drops off differently.
Polished with context
The figure we reviewed last week relates to churn: each cohort drops off at a different rate.
Drag the divider, or use the arrow keys.
1.2× realtime minutes while on
Turbo mode
For a heavy accent or a noisy, distant room, when captions come out in fragments you can't read. AI rewrites the fragments into complete sentences, and every sentence opens to show the captions it came from.
Live captions
so the the electron is所以，这个这个电子是
not uh affected by不，呃，受到影响
the potential at all so这个势，完全，所以
Rewritten
The electron is not affected by the potential at all.电子完全不受这个势的影响。
Some of what was said may be left out or come out inexact. Uses realtime minutes at 1.5× while on, and AI polish pauses.
Conversation
Hold to talk, and they hear it in their own language
About a second after you stop, the other side hears what you said in their language. Or leave both microphones on and just talk.
Say one sentence first — the first voice heard is taken to be you, and other people in the room are ignored. If it picks the wrong person, tap “That’s not me”.
Conversation · hold to talkGuest online
Let's start with where the numbers landed last quarter.
先从上个季度的数据说起。
Retention held, but new signups fell in two markets.
留存稳住了，但有两个市场的新增在下滑。
The guest keeps a text record of this conversation.
Teams
Create a team, invite members, share transcripts
A project can share its transcript with a team, and access follows the project’s sharing setting. Recordings upload only when teammates need to play the audio.
Three roles: owner, admin, member. Ownership can be transferred, and leaving a team unshares the sessions you shared with it.
Teams · shared project0 members
YouOwner
Invited memberAdmin
Invited memberMember
Product weekly · EN to ZHOnly you
Invitation created, email queued.
Session summary
Turn the whole session into notes
Seven note styles, from detailed notes to a bilingual glossary. Once they are written you can ask the notes questions.
A long session has its transcript trimmed before summarizing, and you are told when that happens — nothing is dropped quietly.
Session summary · DeepSeekWaiting
Detailed notes
Concise digest
Academic
Meeting minutes
Action items
Exam prep
Bilingual glossary
Structure preview
Ask the notes a question once they are ready.
Glossary
Pin the terms and spellings for this session
One per line: eigenvector = 特征向量, or just the term on its own. The recognition boost needs a concrete source language, not auto-detect.
Once a term lands, every later caption in this session uses the translation you chose — no drifting between wordings.
Glossary · active this session
00:41From glossary
Then we look at the leading eigenvector.
然后我们看主特征向量。
A glossary applies to this session only.
Start now. Try 90 minutes free.
Recordings stay in this browser unless you choose to upload them. The server stores transcript text and timeline metadata.