Local AI, private and on-premise
AI that runs on your own systems. Not in a cloud.
Tolkr™ is local AI. The models that listen, translate and speak run on your organisation's own systems. No cloud service hears your sound, bills you for it or makes you wait for it.
A cloud service
- Your PC
- The internet
- Their servers
- The internet
- Your PC
Your sound travels, can wait in a queue, and is paid for by use.
Tolkr™, local AI
- Listen
- Understand
- Translate
- Speak
All four steps on your own systems. Nothing travels.
Why local
Three reasons it runs on your own systems.
- No cloud costNo meter running.
No cloud service charges you per minute, per word or per request. Run it all day, or through a whole archive, on the systems you already have.
The cost - PrivacyYour sound stays yours.
Speech, text and translations stay on your systems. During use, no connection leaves the PC.
The privacy - Low latencyNo trip to a server.
No upload, no queue, no download. Captions and a natural voice follow the speaker as they talk.
The speed
Privacy
Private by design, on your own systems.
Interpreting is often about things that must not leave the room: an interview, a board meeting, a source, a call with a partner. With local AI they do not have to.
- Nothing sent during useWe watched every connection during a live session, a dub and a voice test: 43 connections, all inside the PC.
- No accountNo account is needed to translate, and nothing about your use is reported to us.
- Your files, your foldersSessions and dubs are ordinary files in folders you choose on your own systems.
- One choice you makeThe only text that can leave is meeting captions for everyone, when you turn them on.
Low latency
Live, as it is spoken.
Live translation and a live voice depend on speed. With local AI the time goes into the work itself, not into the trip to a server and back.
- 01The speaker talks
Tolkr™ hears the sound as it plays, from the PC or a sound input.
- 02The caption
The original line appears as it is recognised, on your own systems.
- 03The translation
The translated line appears under it, in the context of what came before.
- 04The voice
A natural voice speaks the translation as the speaker talks.
- No queueYour systems only
No upload and no download. Your systems do not wait for anyone else's servers.
- All dayLong sessions
Follow a channel or a stream for as long as it runs, and keep every line.
- Keeps paceBy itself
A fast speaker is followed all the way, and every line stays in the captions.
No cloud cost
No cloud bill. However much you use it.
A cloud translator is usually paid for by use: by the minute of sound, the word or the request. Local AI uses the systems you already have.
- A day of live channels
- No cloud fee
- An archive of recordings
- No cloud fee
- Every meeting of the week
- No cloud fee
- A second language pair
- No cloud fee
What it runs on
Your own systems, sized to it.
- Checks each workstationAt the start
Tolkr™ sets itself up for the workstation.
- Works offlineOnce installed
It listens, translates and speaks with no internet connection.
- Each workstation by itselfNo server
Each workstation runs Tolkr™ by itself, with no shared server to run.
Asked first.
Not answered here? Ask the people who build it
What does local AI mean?
The AI models run on your own systems. Recognition, translation and the voice happen there, not on a cloud service.
Is anything sent to a cloud while it works?
No sound and no text. We watched the connections Tolkr™ opened during a live session, a dub and a voice test, and none left the PC. The only text that can leave is meeting captions for everyone, when you turn them on.
Is there a cost per minute?
Not for the AI. No cloud service is paid per minute, per word or per request, because none is used.
Is it live?
Yes. Captions and a natural voice follow the speaker as they talk. We show it on your own workstation in a demo.
Do I need a graphics card?
It is faster with one, and it also runs on the processor.
See local AI on your own systems.
We run Tolkr™ on a recording or a stream of yours, so you judge it on your own work.
Request a demo