Translate speech to speech, or just transcribe — you chooseAnd when it ends, take the notes with you.
Nobody stops or waits their turn: you finish a sentence and the others hear it in their language a second later. And when the meeting ends, you get the document with what was decided and who takes each task. Out of the same minute: no separate plan, no new bill.
Liora Translation is published in Microsoft's commercial marketplace, and the Teams subscription is billed by Microsoft.
- Voice is never stored
- Fifteen days, then it deletes itself
- Encrypted on both sides
Which one of these are you?
Everyone arrives with a different problem, so this page does not try to sell you all of it at once. Pick the one that sounds like yours; the rest of the site follows from there.
Your room translated, and the notes when it ends
A conference, a convention, a daily service or a council session. We come, we wire the sound, we set up the rooms and an app with your name, and we are there. It stays installed or it gets packed up at the end: you talk, we handle the rest.
- Transcript
- Summary
- Full record
- Everyone hears it in their own language
- No pauses and no taking turns
- Up to fourteen languages at once
Already running in the Cathedral of Santiago de Compostela
See how we set it up
- Someone speaks in their language
- One second
- Everyone else hears it in theirs
All your communication, translated with 
Liora Flow
Messaging, calls, video calls, groups and scheduled meetings: everything you expect from a communication app, with the translation built in rather than bolted on. Everyone writes and speaks in their own, and hears and reads everyone else in theirs. For desktop and phone.
- Voice is never stored
- Fifteen days, then it deletes itself
- Encrypted on both sides
- Transcript
- Summary
- Full record
Translate your Teams meetings, and keep the notes
It is added to the meeting in seconds, a side panel opens, and each person picks the language they want to hear it in. Headphones needed; nothing to install.
- Transcript
- Summary
- Full record
Put real-time audio through it
Audio in over SRT or WebRTC, translated voice out into your own WebRTC over WHIP, and the text by HTTP. The same engine that runs a conference, wired into yours.
See the protocolsBuild with the API
Transcription, translation, speech and the whole speech-to-speech path in a single call. And the rooms your live translation runs in: create one, get the credentials its device connects with, and decide who may take the floor.
Read the referenceDeploy it on your own infrastructure
Every model runs on your hardware: the recognition, the translation and the voice. Not one call to somebody else's API, and it works on a network with no way out.
See what it takesAnd when it ends, choose what you get
Three options, and all three come out of the same minutes: no separate plan, no new bill.
Just the transcript
What was said, written down, and nothing more. No language model in the way — for whoever has to file it and wants nothing interpreting it.
The summary
What was decided and who takes each task, each with the minute it was said. It is what most people want.
The summary with the full record
That, plus the whole document of who said what, punctuated, start to finish. For when the minutes have to exist in full.
Besides translating
Translating is what happens while people speak. This is what happens around it, and it comes out of the same minutes: no separate plan, no new bill.
The notes, when it ends
A document with what was decided, who takes each task and the minute it was said. It joins the queue when the room closes and reaches you by email.
The words of your house
The names only used here: people, streets, references. Written once, so the engine stops swapping them for others that sound similar.
A button if something goes wrong
If the machine translates something it shouldn't, one tap says so. It works without an account, and neither the surrounding conversation nor the audio is sent.
And where does what we say end up?
The question everyone asks before putting this into their company, answered here and not in the small print.
Voice is never stored
Calls, meetings and rooms: they are translated as people speak and then they are gone. The audio is never stored. The text only if you turn it on, and then it is fifteen days and after that it is destroyed.
Fifteen days, then it deletes itself
Everything you write or upload in the app — messages, files, a meeting's chat — is kept encrypted for fifteen days and then destroys itself. One period, no exceptions.
Encrypted on both sides
Our database and the one on your phone and your computer. A lost phone is not a lost conversation.
Nobody else in the path
Hearing, translating and speaking: the whole engine runs on our own machines. There is no outside AI provider to call, so your voice never passes through anybody else's cloud.
The question almost nobody can answer
“Can I keep it inside my own network, with the audio never leaving?” With us the answer is yes: the whole engine — recognition, translation and voice — runs on your hardware, and there is no third-party API to trust because there is none to call. That does not rest on a policy someone can change tomorrow: it rests on where the server is.
How it is deployed in your own houseIt is not end-to-end encryption and we do not call it that: to translate, something has to listen. That period is the app's — in Teams and in the integrations there is nothing to keep. Read it in full in the privacy policy
Measured, not promised.
Three figures from the engine itself, the hardest room we have ever put it in, and the brakes that keep a long session standing.
4 demos, one per language. Open whichever you like.
Non-native speaker (German student) — all other demos feature native speakers.
E2E Latency
The whole path, from someone speaking to being heard translated. Under a second, which is the line past which a conversation stops flowing.
Per minute of translated voice
The same price whichever language it is, and no variable bill from somebody else's cloud. It drops with volume — the detail is on the pricing page.
Languages at once
14 languages in parallel on a single node, out of one listen to the audio, with none queued behind another.
From theory to extreme reverberation.
We deployed Liora Flow at the Cathedral of Santiago de Compostela, one of the most hostile acoustic environments in the world.
Extreme Acoustics
Stone, a vaulted roof and a very long reverberation tail: the hardest place we have put it to work.
Critical Sessions
A whole service, straight through, with nothing restarted along the way.
Data Sovereignty
Local processing: the audio is transcribed and translated inside the premises and does not leave them. Whoever owns the machine owns the data.
Industrial-Grade Security & Resilience
Designed to never fail. Intelligent Circuit Breakers.
Memory Protection: Active monitoring that prevents memory saturation, with an operational ceiling that keeps the system stable however long a session runs.
CPU Load Management: Dynamic load balancing across threads with automatic optimization to keep the system always responsive.
Hardware Isolation: Memory held with margin to spare at peak load, so an unexpected burst of traffic can never exhaust it.
One engine, many uses.
The same acoustic path serves all of it. What changes is whether somebody is speaking right now or the recording already exists.
Real time
Somebody is speaking and the room has to hear it now. There is an hour, there is a place, and it cannot be done again.
- The Liora Flow AppThe whole communication app — chat, calls and meetings — with the translation inside.
- Turnkey eventsA one-off event or an installation that stays: the translation is on us.
- Churches and cathedralsThe Mass, the pilgrims and everything else said the rest of the year.
- Live integrationsBroadcast a service, a talk or a class and let every listener hear it in their own language.
- Microsoft TeamsThe same thing inside a Teams meeting, with nothing to install.
On files
The file already exists and nobody is waiting in a room. Upload what you have and pick the result up.
- Meeting notesWhat was decided, who takes each task, and the minute it was said
- TranscriptionA recording becomes text in the language it was spoken in.
- SubtitlesSRT and VTT with their timings, ready to drop on a video.
- Translate recordingsThat same recording, in every language you need.
- Translate documentsPDF, Word and six more formats, translated whole.
- Who said whatWho said what, marked in the transcript.
Where do you want it to run?
This is not who you are — that was further up. This is where the engine lives once you know what you are buying: our hardware, yours, or one we bring to the door.
Liora Cloud API
Developers, Startups and SaaS applications.
Integrate the fastest translation engine on the market in minutes. Our infrastructure manages the hardware so you can focus on user experience.
Model: Pay per use (minutes processed).
Advantage: Instant elastic scalability with no upfront hardware investment.
Latency: Optimized through high-priority bursts.
On-Premise License
Large corporations, Legal, Healthcare and Government.
Total data control. We install Liora Flow directly on your local infrastructure or private cloud via optimized Docker containers.
Model: Annual recurring license (SaaS On-Prem).
Advantage: Absolute sovereignty. Data never leaves your local network (LAN). Native GDPR compliance.
Requirement: Servers with hardware acceleration.
Hardware Appliance (Liora Box)
Live events, Conference centers and Mobile units.
The turnkey solution. We deliver a pre-configured physical server specifically optimized for the Liora Flow engine.
Model: Hardware sale + Software license.
Advantage: Plug & Play. Connect audio and start processing. No configurations or dependencies.
Hardware: Optimized workstations, supplied ready to run.
Six ways to start.
There is no single form any more. Buying a Teams subscription, asking a question about a protocol and commissioning a deployment are three different conversations, and each one gets its own.
Tell us about your event
A conference, a convention or a cathedral broadcasting daily. Four details — what it is, when, how many people and which languages — and we tell you whether we can do it and what it costs.
Tell us about your eventInstall Liora Flow and talk
Not a translator bolted onto another tool: it is the app your people write, call and meet in, and the language stops being an issue. The download is not open yet, but the page says what it does and where it will run.
See the appSubscribe in Teams
The subscription is bought in Microsoft's marketplace, where your organisation already buys. No form, and no waiting for us to answer.
Microsoft MarketplaceOr read the Teams pageTalk to engineering
Four fields: who you are and what you have in front of you. No monthly volume asked, because at this point nobody knows it yet.
Write to engineeringOr see the protocols firstRead it and start
The account is opened in the client portal and the key is issued from there. The reference is open, the examples run as pasted, and the spec downloads so you can generate a client from it.
From €0.20 a minute of speech
Create an accountOr read the referenceTell us about your project
The long form, and here it is the right one: a deployment inside your own network is decided on volume, on how strictly the data has to stay, and on when.
Open the formOr read the page first