Download MeetVoice
Version 2.0.9
Quick Start
1
Install the app and Chrome extension
2
Start the free trial or enter your license key
3
Start translating — speech recognition is included
Launch the app and start the free trial — 30 minutes of real-time translation, no license required.
Already have a license? Enter your key on first launch.
What's New
v2.0.9
- Pro now costs €19 a month. New subscriptions are billed monthly only; existing yearly subscriptions keep working.
- New Max plan, €39 a month: 30 hours of translation and studio voices for the voiceover in both directions, in 27 languages. Studio voices need MeetVoice Desktop 2.0.9 or later.
- On Max, you choose a studio voice in the extension and can preview it; the extension and MeetVoice Desktop show which voice is speaking and, when the standard voice takes over, why.
- Voiceover is more dependable (with MeetVoice Desktop 2.0.9): since about October 10 it could go silent when the main voice was unavailable, because the backup voice had stopped working — it works again. If the main voice is unavailable when a call starts, it now comes back as soon as it recovers instead of leaving the backup voice until the end of the call, and voiceover no longer switches to the backup voice toward the end of the month.
- A brief drop in speech recognition (up to 10 seconds) no longer shows the “Server disconnected” banner: translation picks up on its own, and with “Only the translation goes to the call” turned on, the other side hears your own voice in the meantime.
- Smaller fixes in MeetVoice Desktop: turning translation back on right after turning it off no longer leaves it off, and when another program is using the app's network port, MeetVoice says so instead of reporting repeated crashes.
v2.0.8
- Translation hours on paid plans now renew every month on the day and at the time you subscribed, instead of on the 1st of each calendar month; yearly plans get their monthly hours on the same schedule. Subscriptions bought before October 8, 2026 switch to the new schedule at the first monthly renewal date for their hours on or after October 8; until then, their hours keep resetting on the 1st of the month.
- MeetVoice Desktop shows the exact date and time your hours renew, in your local time.
- Desktop app: updated Electron engine with the latest Chromium security fixes.
v2.0.7
- The glossary is easier to fill in. “Do not translate” is now a list of names: type a name and press Enter, remove it with the cross. “Translate as” is a list of rows with a term and its translation, so you no longer type the arrow yourself; you can also paste two columns from a spreadsheet or lines such as “term → translation” (→, =>, ->, =, a tab or a spaced dash all work as the separator). Messages about the limit, a repeated entry or a missing translation appear next to the field, a failed save is reported instead of passing silently, and a pair written as “a => b” is no longer saved incorrectly. Area chips such as “Finance” or “Legal” now toggle — click one again to remove it — and a glossary edit made just before you start translating is included in that session.
- Select a word or phrase in the transcript and an “Explain” button appears right at the selection; the explanation opens in a card at the bottom of the panel. The card offers “Do not translate”, which adds the term to your names, and “Set translation”, a small form that adds or updates a “Translate as” pair. Selecting text in the transcript now also works on Google Meet pages, where Meet's own styles used to block it and “Explain” could only show a hint.
- The transcript panel shows what is going on: a loading state, “Phrases will appear here during translation” while there is nothing yet, and “Couldn't load history” with a “Retry” button if loading fails. The panel has a title and a label for screen readers, Esc closes the card first and then the panel, the panel's notices appear inside it, and MeetVoice error notices on the Meet page no longer end up under the open panel.
- Speech recognition is more dependable (with MeetVoice Desktop 2.0.7): when translation starts or recognition reconnects, the last second and a half of what you said while it was connecting is kept instead of being lost; a connection to recognition that hangs while opening is dropped after about ten seconds and retried automatically, instead of leaving translation silent until you press Stop; and a long mute in Google Meet no longer makes the app reconnect recognition every minute, which could end outgoing translation with an error — recognition now resumes when you unmute.
- Voiceover: muting it in the extension window or with Alt+M now also works while a session is still connecting, and it stays muted if you then close the extension window. A voice gender change made while connecting reaches the session too; the voice choice and the voiceover mode are locked for those few seconds and unlock once the session connects or fails to start. A network hiccup while a phrase is being voiced no longer freezes the voiceover — within about five seconds the phrase is voiced by the backup voice or skipped — a slow voice service no longer adds about ten seconds to every phrase, and stopping a voice preview while it is still loading clears the spinner.
- Long translated phrases spoken into the call are no longer cut off after 15 seconds.
- Speaker names stay correct in long meetings: the audio sent to speech recognition no longer slowly runs ahead of the call's audio.
- Japanese, Korean and Chinese text now wraps between words and phrases instead of mid-word in subtitles, the transcript panel and History, and Chinese in subtitles and the panel is shown in a Chinese font instead of a mix of Japanese and Chinese character forms. History uses the right plural forms for the number of entries, shows durations in your interface language and no longer shows a count for an empty meeting. Serbian dates and durations in the extension use Latin script, and the Korean interface no longer addresses you as «당신» in status lines.
- In Markdown and Notion exports, participant names and transcript text are escaped, so a name such as “[Pay](https://…)” stays plain text instead of turning into a link.
- If MeetVoice Desktop is closed or unreachable for several minutes during a call, translation now stops with a clear message that the connection to the server was lost, and you can start it again as soon as the app is back, without closing the Meet tab. After you reload the Meet tab during a session, the transcript panel shows the earlier lines again, and a meeting whose language pair you changed during the call keeps its final pair in History, exports and summaries however it ends. When speech recognition is lost, the disconnected indicator and the error notice now stay on screen, and Alt+S and the volume sliders respond right after the browser wakes the extension.
- Desktop app: installing an update is more dependable. A double click on “Restart to update” no longer opens two confirmations or starts two installs, translation cannot be turned on while the app is closing for the update, and if an install does not finish, MeetVoice comes back in a working state and shows the update error. You will notice this starting with the update to 2.0.7: an update is installed by the version you are updating from, and these fixes are already in 2.0.6. A downloaded update now stays ready: later checks, even offline ones, no longer discard it or announce it again. Closing the license window or choosing the free trial while a key is being activated no longer locks you out, a late reply from the license server can no longer erase a key you have just activated, and restarting translation from the desktop window right after starting it no longer leaves it stopped.
- Desktop app: for extra protection against tampering, it no longer reads the NODE_OPTIONS and NODE_EXTRA_CA_CERTS environment variables. If your company network relied on them to trust its own security certificate, MeetVoice may be unable to connect there.
- Smaller fixes: in multilingual meetings, a short word that recognition cannot assign to a language (such as “okay” between German and Russian) no longer leaves the phrase before it untranslated and unvoiced; a long glossary with names in scripts such as Thai, Arabic, Hebrew or Bengali no longer stops speech recognition from starting; and a phrase that ends with a number, a time or an amount no longer waits five to six seconds for its last words (in English, such numbers may now appear spelled out).