A fast writing rhythm
The core loop is simple enough to become muscle memory: hold a hotkey, speak, release, continue typing.
OiPer desktop / issue 01
A privacy-first voice tool for people who do not want their desktop workflow rearranged around transcription.
Editorial product spread
benchmark leader
1.5s
The promise
Your voice stays yours. The speed stays yours too.
Workflow in one sentence
Hold a global hotkey to capture audio, release to transcribe, and inject the result into the app already in front of you.
A native performance profile
OiPer is written in native code, keeps latency low, and uses GPU acceleration when the machine supports it.
A clear privacy stance
Speech data stays on-device by default. Online optimization only appears when the user deliberately enables it.
Benchmark spread
Speed without leaving the machine behind
Measured on a 30-second English clip. OiPer finishes first while keeping the default path local.
OiPer Desktop1.5s
Lemonfox API3.27s
Python Faster-Whisper3.55s
OpenAI Whisper 1 API6.46s
privacy and optional services
Local processing
Transcription, activity logs, and captured audio remain on the device. That is the baseline, not the premium tier.
Online optimization
Optional cleanup uses your API key, your provider choice, and your chosen model. It can be turned off at any time.
advanced accuracy
When technical language matters, route through an LLM
OiPer supports transcription through LLMs for stronger handling of proper nouns, engineering terms, medical language, or other specialized vocabulary. Lightweight options such as Gemini 2.5 Flash Lite are useful here.
settings and configuration
Speech model downloads and selection by size or purpose
Backend choice for auto, CPU-only, or GPU acceleration
Provider setup with custom base URL, API key, and model name
Text optimization through local processing or online cleanup
LLM transcription mode for technical or specialized language