Polyglot Broadcast transcribes you as you speak, translates on the fly, and re-voices it in your own cloned voice with the right language and accent - Spanish, French, Hindi, Mandarin, Japanese, and more. Your audience picks a language and hears you about two seconds behind live, with subtitles that stay in sync. Available now as a standalone product, and included with the Virtual Call-Center for self-hosted partners.
What it does
You talk normally. Everyone else hears you in the language they chose, in your voice and their accent, close enough to live that it feels like you are speaking directly to them.
You speak in your own language. The system transcribes, translates, and re-voices it into every language your audience is listening in, at the same time.
Clone the speaker's voice once, with their consent, and every translation is spoken in that voice - not a stranger's. It still sounds like you in French, Japanese, or Hindi.
British English, American English, and more are distinct choices. Each listener hears the register that fits their region, not one flat machine voice.
Listeners tap one toggle to switch languages and the audio follows instantly, with no gap and no reconnect. Only the languages people are listening to get synthesized.
Drive a channel's language from a timeline: by the clock, by paragraph, or by a spoken cue. Open a keynote in French, move to English at the ten-minute mark, switch on "moving to Q and A."
Every phrase carries its own timing, so captions land with the audio, not before or after. Read along or turn them off - either way it matches.
How it works
Five stages, streamed the whole way. Nothing waits for you to finish a sentence - each phrase is chunked, translated, and voiced as it lands.
Streaming speech-to-text turns your voice into text with word timing and clean phrase breaks.
Phrases and paragraph breaks are detected so the system can start work before you finish talking.
Each active language is translated on the fly, tuned to the right regional register.
Streaming text-to-speech speaks each translation in your cloned voice and the target accent.
Audio and matched subtitles reach every listener on the language track they chose.
A closer look
On a live channel, every listener picks their own language from a toggle. Switching is instant - the track they hear changes with no reconnect and no gap - and only the languages people are actually listening to are synthesized, so cost and latency stay bounded.
A programmed channel picks the output language for you, from rules you set: by the clock, by which paragraph you have reached, or by a spoken cue. Perfect for a scripted keynote where the language is supposed to change on schedule, hands-free.
Getting started
Start a free broadcast in your browser, pick the plan that fits, and go live. Already a self-hosted partner? It is already part of your Virtual Call-Center floor.
Record a short consented sample of the speaker once. That voice becomes the one every translation is spoken in.
Pick the languages and accents your audience needs, or set a programmed timeline that switches them for you.
Start speaking. Listeners choose a language and hear you about two seconds behind, subtitles in sync.
Pricing
You pay for language-minutes - one minute heard in one language. A 30-minute talk heard in 4 languages is 120 language-minutes. Only the languages someone is actually listening to get synthesized, so you never pay for an empty channel. Every plan starts with a free trial.
Webinars and smaller town halls.
Regular events and multilingual town halls.
Conferences and large live audiences.
Prices in US dollars. Already a WorldVC self-hosted partner? Polyglot Broadcast is included with the Virtual Call-Center.
Better together
Every WorldVC app works alone and works better together. Broadcast is how the AI floor reaches an audience in their own language, live.
The AI-run call center where Broadcast lives today, so every caller can be heard in their own language.
See Virtual Call-Center →The voice stack behind it all - the same real-time speech pipeline, on a single line.
See VocalizeCloud →Line up a multilingual briefing or town hall and let attendees book their seat.
See Scheduler →Questions
Yes. It is available now as a standalone product in three plans - Starter at $99/mo, Pro at $249/mo, and Business at $599/mo - each with a block of language-minutes included and metered usage after that. It is also included with the Virtual Call-Center for WorldVC self-hosted partners.
Each plan is a flat monthly base that includes a block of language-minutes - one language-minute is one minute heard in one language. Go past your included block and extra language-minutes meter at your plan's rate ($0.40, $0.30, or $0.20). Because only the languages someone is actually listening to are synthesized, you never pay for a channel no one is on. All prices are in US dollars.
Yes. Every plan starts with a free trial: you can broadcast free up to a set number of minutes, and once you reach the limit you are prompted to subscribe to keep going. Nothing is deleted - you just pick up where you left off after subscribing.
Yes. With the speaker's consent, we clone their voice once, then every translation is spoken in that voice with the target-language accent. Consent is a hard requirement and is recorded for each broadcast.
Roughly two to three seconds from the moment you speak to the moment a listener hears the translation, because each phrase is transcribed, translated, and voiced as a stream rather than waiting for you to finish a sentence.
On a live channel each listener taps a toggle to pick their language, and the audio switches with no gap or reconnect. On a programmed channel you set the language yourself with rules based on time, paragraph, or a spoken cue.
The pipeline handles the major world languages plus regional accents - for example British English and American English as distinct choices. Tell us the languages your audience needs and we will confirm coverage.
Start a free broadcast now, then pick a plan from $99/mo. Also included with the Virtual Call-Center for self-hosted partners.