Lven Instant — on-device, everywhere
Voice-to-text that runs entirely on your machine.
Lven started as a question about friction: why does voice input require so many steps? Open an app, record, wait, copy, switch windows, paste. Six steps for something that should be one.
The answer is a floating bubble. It sits on your screen, appears when you are in a text field, and disappears when you are not. Tap it, speak, and the text lands at your cursor. No clipboard, no paste, no app switching.
Lven Instant takes that idea and removes the next compromise: privacy. Every word you dictate is converted to text by a quantised speech model that ships inside the app and runs on your CPU. No upload, no round trip, no remote inference. The audio physically does not leave your device.
The model runs at usable speed on five-year-old hardware. Air-gapped machines transcribe exactly the same as connected ones — minus the periodic check that confirms the subscription is paid. That check carries your account token and nothing else.
Three tiers: 6 months at $29, 1 year at $49, or a one-time $139 lifetime. One subscription covers Windows, macOS, Linux, Android, and iPhone. The original cloud build remains available at /releases/lven/cloud for users who specifically want server-side transcription.