OiPer Desktop transcribes speech to text at native speed — entirely on your device. Hold a hotkey, speak, release. Done in 1.5 seconds.
A workflow that becomes invisible. Hold, speak, release — and you're done.
Native code. GPU acceleration. Results in 1.5 seconds. Not a round-trip.
No telemetry. No cloud by default. Your audio stays exactly where it should.
Six capabilities that work in concert to make voice transcription feel effortless.
Hold any global hotkey to begin recording. Release to transcribe. An interaction so natural it disappears.
Whisper runs on your own hardware. Your voice, your words, your machine — never surrendered to a server.
Transcribed text arrives in your active application. No clipboard, no paste, no interruption.
CUDA and Metal support built in. Leverage your graphics card for inference that feels instantaneous.
Engage any LLM — local or cloud — for text cleanup. Always opt-in, always under your control.
Choose your Whisper model. Tiny for speed, Large for precision. Download and switch from within the app.
OiPer processes 30 seconds of English audio in just 1.5 seconds — more than twice as fast as the closest API competitor, while keeping your data entirely on-device.
Your data is not a product. OiPer is designed from the ground up to keep everything on your device.
Whisper runs on your own CPU or GPU. No model API, no server, no exposure of any kind.
Activity history and raw audio are stored locally. You choose what to keep and what to delete.
Text cleanup via external LLMs is entirely opt-in. Supply your own API key. Disable anytime.
No subscription. No cloud dependency. A single download and your voice becomes your fastest input.