Vays turns your voice into finished text, summaries and action items. All of it runs on your Mac. No cloud, no account, no subscription. You buy it once.
WHAT VAYS DOES
• Records voice memos directly on your Mac
• Transcribes on device, using Apple's speech engine or WhisperKit models
• Tells speakers apart and attributes what was said to the right person
• Translates live from Polish to English, with captions ready for OBS
• Runs a presenter mode so you never lose your key points mid-talk
• Cleans raw transcripts into readable text
• Cuts long recordings down to what actually matters
• Pulls out tasks, decisions and follow-ups
• Lets you chat with your recordings using a local language model
100% LOCAL, 100% PRIVATE
Every transcription, every summary and every AI response is computed on your Mac. No cloud round-trip. No telemetry. No account to create. Your memos stay on your machine. Period.
BUILT FOR APPLE SILICON
Vays runs Core ML models with Metal acceleration, tuned for M-series chips. Transcription is fast, summaries feel instant, and a local language model runs comfortably on an M1 or later.
POWERFUL WHEN YOU NEED IT, QUIET WHEN YOU DON'T
• Quick Tasks: thirty seconds of talking becomes a structured to-do list
• Long recordings: meetings and interviews get transcribed, summarised and indexed
• Chat with a memo: ask what was said and get an answer
• Calendar context: see what is relevant today
• Search: go through your whole voice archive as if it were text
FOR PEOPLE WHO TALK FASTER THAN THEY TYPE
Vays was built for construction managers, field engineers, researchers, doctors and consultants. For anyone who captures thoughts out loud. If you have ever lost an idea because typing broke the flow, this is for you.
WHAT YOU GET
• Unlimited recording
• On-device transcription (Apple engine and WhisperKit)
• Speaker recognition with a voice library
• Local chat and summarisation
• Transcript cleanup engine
• Live translation for streaming
• Presenter mode
• Calendar context
• Export to text, Markdown or the clipboard
REQUIREMENTS
• macOS 14 or later
• Apple Silicon recommended
• Disk space depends on the models you choose. Apple's speech engine downloads nothing. WhisperKit models range from roughly 150 MB to 1.5 GB. A local language model is typically 2 to 5 GB.
PRIVACY
Vays does not collect, transmit or store any of your data off device. No analytics. No crash reporters phoning home. The only network calls Vays makes are to localhost, so the app can talk to its own local inference server.
By Arvind Juneja.
Support on Discord, link inside the app.