It figures out the rest.
Speak, and the right language is detected automatically — no picker, no setting, no restart. Three engines behind it, each tuned for something different.
Speak in any language.
20+ languages with automatic detection — switch mid-sentence and it keeps up. No setting to change, no language to pick before you start talking.
Every language above, fully offline, on-device.
Tuned specifically for Hindi, Tamil, Bengali, Punjabi and mixed Hinglish.
The most accurate option for tricky accents and fast speech.
Your voice stays on your machine. Your brain stays in your cloud.
Dictation and speaker separation run on-device. Summaries and answers go to a model that is told not to train on them. And for organisations, the whole stack deploys inside your own cloud account — we run the software, we don't hold the data.
On-device Whisper
PrivateZero network. Works on a plane. Nothing leaves.
- Apple Silicon and Windows GPUs
- Speaker separation on-device too
- Free, no account needed
OpenAI or Sarvam
Your keyBest-in-class accuracy, or Indic-tuned models for Hindi, Tamil, Bengali and Hinglish.
- Bring your own API key
- Per-language engine choice
- Switch any time
Notes, answers, actions
Our accountNothing is sent until you ask. Every AI output is a draft you can edit.
- Providers configured not to train
- Counts recorded, never content
- Read what's sent, and when
One deployment per organisation, inside your account. Meetings, embeddings, connector tokens and models never leave it. We operate and upgrade it.
Native on every desk and in every pocket.
Download Kept for free. No account required, no API key, no setup — just install and start talking.
Open source · Sign in with Auth0 only when you want meetings synced, shared or asked about.