Chrome extension · Real-time meeting translation

Speak any language.Be heard in theirs.

Live translated captions and an AI voice that speaks for you — inside Teams, Meet, Zoom, and Webex. One click to start, 60 languages, and your meeting audio never touches our servers.

Be first in line when Gaavala lands on the Chrome Web Store.

We'll email you about the launch — nothing else. One-click unsubscribe. Privacy Policy

Gaavala

Live
Source: auto-detect
00:26YukiJA → EN

金曜日までにレビューを終えられます。

We can finish the review by Friday.

00:33CarlosES → EN

Necesito un día más para el informe de facturación.

I need one more day for the billing report.

00:40AnnaDE → EN

Das Budget ist freigegeben wir können starten.

The budget is approved — we're clear to start.

00:47LéaFR → EN

Parfait. Je préviens le client cet après-midi.

Perfect. I'll let the client know this afternoon.

languages
60
platforms, one install
4
studio voices
69
overlay refresh
100ms
Two colleagues in a sunlit room, on a video call with a remote participant

Sixty languages · detected per speaker, on the fly

EnglishTürkçeDeutschFrançaisEspañolPortuguêsItalianoNederlandsPolskiРусскийУкраїнськаالعربيةעבריתहिन्दीবাংলাதமிழ்中文日本語한국어ไทยTiếng ViệtBahasa IndonesiaBahasa MelayuFilipinoSvenskaNorskDanskSuomiČeštinaSlovenčinaMagyarRomânăБългарскиΕλληνικάHrvatskiSlovenščinaLietuviųLatviešuEestiCatalàفارسیاردوKiswahiliAfrikaans
In a real meeting

Captions live where the meeting lives.

A floating glass layer renders translated captions directly over the call — you never look away from the people talking. The overlay is real product footage; the meeting behind it is one we built, with people who do not exist.

Gaavala's floating subtitle overlay rendering translated captions over a live video meeting, with speaker colors and age-fade.
Fig. 021440 × 810
  • Age-fade stream

    New lines land at full strength, older ones fade back — your eyes always find the newest words.

  • 8 speaker colors

    Every voice gets a consistent color — in the overlay, the side panel, and your exports.

  • Glass over the call

    Blurred backdrop, adjustable font size, opacity, and line count — readable without blocking faces.

Understand

Follow every voice in the room.

A bilingual transcript that keeps the original and its translation together — with speakers named, colored, and exportable.

Transcript

Original + translation, paired

Timestamps, language badges, and both sides of every sentence — what was said and what it means stay together.

[00:14:32] Mehmet TR Bizim taraf hazır.

→ EN Our side is ready.

Language detect

No source language to pick

Gaavala detects each speaker's language on the fly — even when the room keeps switching. You only choose the language you want to read.

ENTRJADEESFRPTKOARHI+50

Diarization

Eight colors, real names

Speakers are detected automatically and color-coded. Rename "Speaker 2" to a real name mid-meeting — captions, overlay, and exports update instantly.

Speaker diarization in the Gaavala side panel: color-coded speakers, with one being renamed live from Speaker 2 to Ayşe Kaya.

Live transcript

Speaker 1

Let's lock the launch date. Can everyone ship by the 18th?

Let's lock the launch date. Can everyone ship by the 18th?

Ayşe Kaya

Bizim taraf hazır — testleri dün gece bitirdik.

Our side is ready — we finished testing last night.

Speaker 3

金曜日までにレビューを終えられます。

We can finish the review by Friday.

Ayşe Kaya

Necesito un día más para el informe de facturación.

I need one more day for the billing report.

Fig. 031440 × 810

Export

TXT, SRT, JSON — translations included

Subtitles for video editors, JSON for your tooling, plain text for everything else — previewed live before you download.

gaavala-transcript-2026-06-10.txt
[00:00:02] Sarah (en → en)
  Let's lock the launch date. Can everyone ship by the 18th?

[00:00:07] Mehmet (tr → en)
  Bizim taraf hazır — testleri dün gece bitirdik.
> Our side is ready — we finished testing last night.

[00:00:12] Yuki (ja → en)
  金曜日までにレビューを終えられます。
> We can finish the review by Friday.
Fig. 04Click a format — or let it cycle

Speak

Talk back — in a voice they understand.

One button: tap it for a hands-free 60-second window, or hold it to talk only while you hold. The meeting hears a natural voice saying it in theirs. No virtual microphone, no platform setup.

  • One button — tap it or hold it

    A tap opens a 60-second hands-free window. Hold instead and you talk only while you hold. Tap again to stop early.

  • The ring counts the window down

    A draining arc counts the window down in minutes and seconds, turning amber for the last 10 seconds. Tap twice within 0.3 seconds and the window becomes 30 seconds.

  • No virtual mic driver

    Your translated voice is grafted into the meeting's own audio sender — and your real mic auto-mutes while it plays, so participants never hear both voices.

  • The platform's mute always wins

    Mute yourself in the meeting and Gaavala's output is silenced with you — your mute reflex stays trustworthy.

One button

Tap it. Say it once.

0:42Left
Tap
60-second hands-free window
Double-tap
30 seconds instead
Hold
talks only while held

You said · EN

“Let's lock the launch date. Can everyone ship by the 18th?”

The meeting hears

Q3 planning● LIVE
Sarah
EN
Mehmet
TR
Yuki
JA
Mehmet

Our side is ready — we finished testing last night.

Yuki

We can finish the review by Friday.

Carlos

I need one more day for the billing report.

You · speaking in EN · via Gaavala

“Let's lock the launch date. Can everyone ship by the 18th?”

no virtual microphone · your mic auto-mutes while it plays

Fig. 05The Speak window counting down, mid-send

NewIn your own voice · Pro

Speak in your own voice.

Record about twenty seconds, once. Gaavala clones your voice and it speaks all 60 languages — so the meeting hears you, not a stand-in.

AES-256-GCM encrypted at rest · delete removes the recording and the synthetic voice.

69 studio voices — or your own.

Every voice speaks all 60 languages — cross-lingual, zero setup, streaming in real time. Pick a studio voice, or clone your own from a 20-second sample.

Studio voices
69
Female
29
Male
40
Languages, every voice
60
Setup
none

Female — 29

  • Arabella
  • Bianca
  • Bonnie
  • Brooke
  • Colleen
  • Cordelia
  • Emma
  • Freya
  • Grace
  • Harper
  • Hazel
  • Imogen
  • Iris
  • Isla
  • Juliet
  • Kayla
  • Margo
  • Mina
  • Nina
  • Nora
  • Piper
  • Poppy
  • Priya
  • Reese
  • Reyna
  • Ruby
  • Sari
  • Sloane
  • Victoria

Male — 40

  • Adrian
  • Alistair
  • Arjun
  • Arthur
  • Bennett
  • Bryce
  • Cooper
  • Curtis
  • Daniel
  • Dominic
  • Elliot
  • Emerson
  • Emilio
  • Evan
  • Freddie
  • Harlan
  • Haruto
  • Hugo
  • Karan
  • Kenji
  • Logan
  • Mason
  • Mateo
  • Miles
  • Nathan
  • Nigel
  • Oliver
  • Owen
  • Preston
  • Rafael
  • Reid
  • Rohan
  • Russell
  • Sebastian
  • Shane
  • Silas
  • Trevor
  • Tunde
  • Walter
  • Wesley

— or clone your own, from one 20-second sample.

Fig. 0669 studio voices · 29 female / 40 male · every voice speaks all 60 languages
A person mid-sentence wearing a headset microphone, in daylight
On-device AI

The meeting summary that never leaves your device.

Chrome's built-in AI model turns the transcript into structured notes — on your hardware, with zero network calls.

0 network calls
Summaries run on Chrome's built-in model (Gemini Nano). The transcript is never uploaded — not to us, not to anyone.
Key Points · Action Items · Decisions
Schema-enforced structure with speaker attribution — "Mehmet owns the billing report", not vague prose.
5 output languages
English, Spanish, Japanese, German, French — picked automatically from your caption language.
Copy, or download as .md
One click into your notes, with localized headings.

Requires Chrome's built-in AI (Gemini Nano) — availability is checked right in the panel

Gaavala's summary panel generating Key Points, Action Items and Decisions from a meeting transcript, on-device.
Fig. 07Key Points · Action Items · Decisions, generated on-device
Hands resting at an open laptop on a desk in daylight
Privacy, with receipts

Don't trust us. Verify us.

Every claim below maps to a specific line of the product's code — the file reference is printed under each one.

  • The API key in your browser dies in 60 seconds

    Sessions run on single-purpose temporary keys minted per use. The master key never leaves the server.

    soniox.service.ts — expires_in_seconds: 60

  • Audio goes browser → Soniox. We're not in the path.

    An encrypted WebSocket streams audio straight to Soniox's EU endpoint. No Gaavala server can see, store, or log it.

    wss://stt-rt.eu.soniox.com — browser → Soniox, EU

  • AI summaries make zero network calls

    Chrome's built-in model summarizes on your hardware, and the session is destroyed after each run.

    prompt-api.ts — window.LanguageModel, 0 network calls

  • Transcripts live in memory, nowhere else

    No database writes, no server persistence. The only copy is the one you export to your own disk.

    in-memory only — no DB write, local export

  • 10 permissions. Zero analytics SDKs.

    Only the four meeting platforms and three API hosts — no <all_urls>, no tracking libraries. Read the manifest yourself.

    manifest.json — 10 permissions, no <all_urls>, 0 analytics SDKs

Encrypted in transit (TLS 1.2+). Speech processing on EU endpoints. Built to align with GDPR and KVKK.

SOC 2: on the roadmap — not claimed until audited

Animated architecture diagram: audio flowing from the browser directly to Soniox in the EU, with Gaavala's servers off the audio path and a 60-second key countdown.
Fig. 08Audio leaves the browser for Soniox — no Gaavala server on the path

Gaavala is heading to the Chrome Web Store. Join the waitlist and you'll get the install link — and the launch discount — the day it's live.

Join waitlist
Platforms & reliability

Four platforms, one install — and it actually stops when you stop it.

Official Speak Mode support and live captions on the big four — and the reliability engineering nobody puts on a landing page.

  • One-click capture — no dialogs

    Press Start and Gaavala captures the meeting tab directly. No screen-share picker, no "which window?" moment.

  • It actually stops

    Close the panel and capture tears down completely — stall watchdogs and keepalives make sure nothing keeps listening in the background.

  • Survives real life

    Headphones die mid-call, the extension updates itself, Teams swaps its audio stack — capture recovers without breaking your meeting.

Pricing

Honest pricing: $19.99 a month, 120 minutes a day.

Start with 5 free minutes — no card. The trial is once-ever, the limits are real, and all of them are on this table.

What you actually getFree$0no card requiredPro$19.99/month
Live caption minutes5 minutes — once, ever120 min / day · resets 00:00 UTC
Languages & speaker colorsAll 60 · 8-color diarizationAll 60 · 8-color diarization
Speak Mode (an AI voice talks for you)Included
Studio voices69 voices · 60 languages
Voice clone (your own voice)Included — 20s sample, 60 languages
On-device AI summaryIncludedIncluded
Transcript exportTXT · SRT · JSONTXT · SRT · JSON
Caption capacity priorityStandardReserved capacity

No "unlimited" anywhere. Pro is 120 minutes of live captioning per day — enough for back-to-back meetings — and the cap is printed here because you'll see it inside the product too.

Gaavala is heading to the Chrome Web Store. Join the waitlist and you'll get the install link — and the launch discount — the day it's live.

Join waitlist

Why Gaavala

  • Real-time translation in 60 languages — not just captions in one.

  • One install across Teams, Meet, Zoom, and Webex.

  • Speak back in your own voice — or 69 studio voices.

  • Privacy-forward: meeting audio streams straight to Soniox and is never stored; your voice clone is encrypted and deletable.

  • One honest price — $19.99/mo, 120 minutes a day. No "unlimited" asterisks.

Billed through Paddle (Merchant of Record). Taxes and invoices handled automatically.

FAQ

Frequently asked questions

Gaavala is a Chrome extension that adds real-time translated captions and an AI speaking voice (Speak Mode) to meetings running in your browser. Join a Microsoft Teams, Zoom, Google Meet, or Webex call, open the Gaavala side panel, and press Start — Gaavala captures the meeting tab's audio in one click, with no screen-share dialog. It detects each speaker's language automatically; you only pick the language you want to read. Audio streams from your browser directly to the Soniox speech engine over an encrypted WebSocket, and live captions appear in the side panel and in a floating overlay on the meeting.

No. Meeting audio flows directly from your browser to Soniox's speech-to-text servers in the EU over an encrypted WebSocket (TLS 1.2+). Our backend is never in the audio path — we cannot see, store, process, or log meeting audio, and sessions run on temporary keys that expire after 60 seconds. The only things our servers handle are authentication (Google or Microsoft sign-in) and your subscription status. AI summaries run entirely on your device via Chrome's built-in model, so the transcript never leaves your machine either.

Officially: Microsoft Teams (including the new Teams web client), Zoom web client, Google Meet, and Cisco Webex — with the full feature set, including the floating caption overlay and Speak Mode voice injection. Desktop apps are not supported; join each meeting from its web client in a Chrome tab.

Speak Mode is Gaavala's AI voice feature, available on Pro. You talk in your own language — one button: tap it for a hands-free 60-second window with a visible countdown, tap it twice quickly for 30 seconds instead, or hold it to talk only while you hold — and Gaavala transcribes, translates, and speaks your sentence into the meeting in a natural voice. You speak as one of 69 Soniox studio voices — every voice covers all 60 languages, zero setup — or clone your own voice from a ~20-second sample and be heard as yourself. If you mainly need to understand others, captions are enough; if you need to be understood, Speak Mode is the reason to upgrade.

No. While Speak Mode plays your translated voice, your real microphone is automatically muted — participants hear only the translation. And the meeting platform's own mute button always wins: if you mute yourself in Teams, Zoom, Meet, or Webex, Gaavala's output is silenced along with you.

The free trial is 5 minutes of live captioning with no credit card — a one-time, lifetime allowance that never resets, intentionally enough to test it against a real meeting segment. Pro is $19.99 per month and includes 120 minutes of live captioning per day (resetting at midnight UTC), Speak Mode with 69 studio voices and your own cloned voice, and reserved capacity priority. There is no "unlimited" tier — 120 minutes a day is the honest number. Billing runs through Paddle, and you can cancel anytime from the customer portal.

Gaavala is a Chrome extension built on Manifest V3, so it runs on Chromium-based desktop browsers — Google Chrome, Microsoft Edge, Brave, Arc, Opera, Vivaldi — version 116 or newer. Safari and Firefox don't share Chrome's extension API surface, and mobile browsers don't support extensions at all. For mobile-heavy days, join the meeting from a desktop Chromium browser and let Gaavala run there.

No tenant-level approval is required. Gaavala runs locally in your browser, signs in with your personal Gaavala account, and talks directly to Soniox — it never touches your Microsoft 365 or Google Workspace tenant. The extension requests 10 permissions scoped to the four meeting platforms and three API hosts — no <all_urls>, no tracking SDKs. The only IT question is whether your organization allows Chrome extensions at all.

Summaries run 100% on your device using Chrome's built-in AI model (Gemini Nano) — zero network calls, and the session is destroyed after each run. You get structured Key Points, Action Items (with the responsible speaker named), and Decisions, downloadable as Markdown. It requires a recent Chrome with the built-in model available on your hardware; Gaavala checks availability in the panel and tells you honestly if your setup can't run it. Output languages currently supported by the on-device model: English, Spanish, Japanese, German, and French.

Yes, on Pro. Record about 20 seconds once; your cloned voice then speaks all 60 languages. It's powered by Soniox — no account or API key to bring.

It's biometric data, treated as such: stored encrypted (AES-256-GCM) on our servers, used only to build and maintain your voice, and processed by Soniox. Nothing else.

Anytime, from the extension. Deleting removes the recording, the synthetic voice, and Soniox-side data.

All 60 supported languages, cross-lingual, from a single ~20-second sample.

A late video call lit by a single desk lamp

Your next meeting could be in any language.

Join the waitlist — get the install link and a launch discount the day Gaavala hits the Chrome Web Store.

We'll email you about the launch — nothing else. One-click unsubscribe. Privacy Policy

No spam · one-click unsubscribe · meeting audio never touches our servers