Deploy it on your own infrastructure.
Liora Flow Sovereign includes the whole engine — recognition, translation and voice — plus the Liora Flow App and web client inside your infrastructure. The audio does not leave your network, and there is no external API to trust, because there is none to call.
What sovereign means here
It is a word everybody uses for something slightly different, so here is ours, in four things you can check rather than four adjectives.
Nothing leaves
Speech recognition, neural translation and voice synthesis all run inside the container, on your host. There is no call out to anybody's API, which is why it also works on a network that has no way out at all.
It is yours to run
It ships as a Docker image with hardware acceleration enabled on the host. Configuration is environment variables and YAML, so it drops into the CI/CD you already have rather than asking you to build a new one.
Compliance by design
When the audio never crosses your perimeter, the data protection question stops being a clause in a contract and becomes a fact about the network. That is a different conversation with your legal team.
The Liora Flow App is yours too
Your organisation can use private Liora Flow App and web clients, customised with your branding and logo, connected to the engine deployed inside your network. Your people can write, call, make video calls and create meetings; in them they can activate real-time translation, subtitles and notes from the web. Meetings are shared by link: external users join from any browser, without installing the App, creating an account or buying Liora Flow.
The complete experience, inside your network
Sovereign is not just the engine: it is the complete Liora Flow experience running on your own infrastructure, with your organisation's branding, logo and access.
Everyday communication
Your people can write, make calls and video calls, create meetings and send audio from private Liora Flow App and web clients. Translation is integrated into the conversation, not added as a separate tool.
Live translation
Each participant chooses their language and receives real-time voice translation. You can also activate subtitles and meeting notes from the web whenever needed. Minutes are spoken minutes: silence is not counted and does not consume them.
File processing
Beyond live sessions, your organisation can transcribe and translate recordings, documents and audio in the same sovereign environment, without sending content to external services.
Organisation control
You decide who creates rooms, who joins, which keys are used, how long content is retained and which logs are kept. The app, web client, name, branding and logo are presented as part of your organisation; guests only need the link.
Your meetings scheduled and in view
Scheduled meetings, contacts and calls in the same place where the talking happens. No stepping out to another program to see what you have today.
How the translation works
Two languages, not one
You choose separately which language you speak and which one you listen in. Two settings, not a pair.
The voice you are translated with
You choose the gender of the voice you are translated with.
Your real voice, underneath
Your real voice plays quietly under the translation, with its own control. You can tell when the other person has finished — and who is speaking.
It is read as well as heard
Subtitles on video calls: what the other person says, and what you said. If it misheard you, you see it there and then.
Nobody is left out by oversight
If somebody has not said which language they speak, Liora Flow spots it and lets you ask them with one tap.
And in a group, everyone in their own
The audio is heard once and every language comes out of that one listen. Nobody has to give way or agree on a common tongue: you speak in yours and hear the others in yours, even if there are four of you and no two match.
When people are talking.
Four ways, and in all of them the translation runs both directions.
Calls and video calls
One to one, with live translation in both directions.
Group calls
Every participant speaks and listens in their own language.
Meetings
The same thing, with more people and scheduled in advance.
In-person translation
Two people, one computer and no call at all: a consultation, an interview, a service desk. Desktop app only, and with headsets whose microphone does not pick up the other person's voice.
The chat gets translated too.
Not just the calls: also what people write to you and what they send you recorded.
Text messages
It arrives already translated, and the original is not lost: it stays on top. The translation is saved with the message — come back a week later and it is still there.
Voice notes
A voice note in another language becomes four things, and you choose which ones to see.
Original
The audio
Original
Transcribed
Translated
The audio
Translated
Transcribed
What gets said lasts only as long as it must
Voice is never stored. What you write or upload, fifteen days — and only so it reaches your devices, encrypted here and encrypted there.
Voice is never stored
Not calls, not video calls, not meetings. They are translated while you speak and gone: a call leaves no audio, no transcript and no translation.
Fifteen days, then it deletes itself
Messages, files and a meeting's chat: the same for all of it. They are kept encrypted on our server for one thing only, to reach your other devices, and after fifteen days they destroy themselves without anyone having to remember.
Your history, on your device
That is where the conversation stays: in the local database on your phone and your computer, encrypted as well. Delete the app and it goes with it.
Nobody reads it or studies it
We do not analyse what you say, we do not train anything on it and we do not hand it to anyone.
It is not end-to-end encryption and we are not going to say it is: to translate, something has to listen, and the message is encrypted on our server with our keys. What is true is what is above — voice is not stored, everything else lasts fifteen days, and what remains, remains on your device.
And around it, what you expect from a communication app
Full messaging
Text, photos, video, documents and voice notes. One to one and in groups.
Meetings without friction
Shared with a link or a six-character code. Waiting room and password, both optional.
Guests with no licence
Anyone not on Liora Flow joins from the browser: nothing to install, no account and nothing to pay.
Your devices at once
The same account on the computer and the phone. Answer on one and the other stops ringing.
15 languages
To speak in and to listen in. The same ones the engine translates into everywhere else.
Those are the languages it translates into. The app itself — its menus and its buttons — is translated into thirteen, and the phone one into sixteen.
A real video call, with the real app. Not a mock-up and not a re-enactment: this is what you hear and what you read.
Translating is optional
A room can translate, subtitle, or do both. What never happens is the same sentence being translated and subtitled: it is translated when there is somebody to translate for, and subtitled when there is not.
Subtitling without translating costs 1.25 credits per spoken minute and does not multiply per language, because what you see is what is said. And it is decided sentence by sentence: turn translation on halfway through and what came before still counts as subtitles.
Subtitles are always seen in the meeting, kept or not. Keeping is a different thing and the account decides it, not the meeting: what is said is kept for fifteen days and then destroyed.
Translate
COMES WITH THE TEXT
Text
TURNS ITSELF ON
Voice and text, and the minutes when you are done.
No text, no minutes: the room isn’t created.
Translating already brings the text
You cannot translate without first recognising what is said, so a room that translates is already producing the text. It is not a separate product: it comes with the translation.
And you can subtitle without translating
That is the council meeting held entirely in one language: there is nobody to translate for, and the text is needed all the same. And also the room that translates and where, today, everybody speaks the same — with both switched on it subtitles while they share a language, and moves to translating on its own the moment somebody else walks in.
What you cannot do is keep it with neither
No text, no minutes, and it is refused rather than leaving you a room that says it keeps and keeps nothing.
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.
Built for Infrastructure
Powered by a proprietary orchestrator engineered for ultra-low latency and massive concurrency. Every component runs on-premise — from speech recognition to neural translation to voice synthesis.
Native Orchestrator
Custom-built runtime for real-time audio pipeline management.
100% On-Premise
Zero external API calls. All models run locally on your infrastructure.
Massive Concurrency
Handle hundreds of simultaneous audio streams with a native concurrency architecture.
Industrial-Grade Security & Resilience
Designed to never fail. Intelligent Circuit Breakers.
An eight-hour session
the brake has acted
MEMORY
CEILING
CPU
CEILING
RESERVED HEADROOM
FLOOR
THE SESSION GOES ON
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.
And when there is no machine room.
The Liora Box is how you get sound into Liora Flow when there is no rack and nobody to set one up: an event, a conference centre, a mobile unit. A ready-configured physical unit that takes the audio from your desk and sends it to be translated.
- It arrives assembled. Connect the audio and it works — there is nothing to configure.
- This is not the deployment on your premises: the Box captures and sends, and the engine translates on our servers. If what you need is for the audio not to leave your network, what you want is the licence above, on your own hardware.
- Sale of the unit plus the service, and the stations are delivered ready to work.
Technical FAQ
This section addresses the integration, security and performance challenges of Liora Flow.
Start Your Project with Liora Flow
Tell us about your technical challenge. Our engineering team will analyze your use case to propose the most efficient inference architecture.
The jargon of your trade
The engine hears everyday language well. It fails on the vocabulary used only inside a profession — «paroxysmal nocturnal dyspnoea», «plea of lis pendens», «tightening torque» — and it always fails the same way: it swaps them for ordinary words that sound similar.
- It is tuned with material from a field, not from one house: paroxysmal dyspnoea is the same in every hospital. That is why shared tuning is not the cheap option, it is the one that works better.
- The glossary comes first, and it is almost always enough. This is the step above, for whoever already tried it and came up short.
- It is not switched on with a checkbox: you talk, you get a quote, you sign. And it switches off with a button, no questions, immediately.
On your own servers you need a card with at least 24 GB: the model does not fit on a desktop one.