hold hotkey then record then release then instant injection

OiPer Desktop is optimized for people who think fast and type faster. It captures speech, transcribes with low latency, and inserts output directly into the active application.

ux: minimal friction

performance: native speed

privacy: local by default

Execution log

[local] audio capture initialized
[local] whisper engine loaded
[privacy] outbound traffic: disabled
[action] release hotkey -> transcription
[result] injected text in active window
[optional] online cleanup only if enabled

Privacy and security policy

- Transcription runs locally on your machine.
- Activity logs and audio remain on-device.
- Online optimization is optional and user-triggered.
- API key is required for online providers.
- Disable online mode at any time from settings.

Config matrix

model: whisper-large-v3 or lightweight local model
backend: auto | cpu | gpu
cleanup: local or online
online: base_url + api_key + model_name
advanced: llm-transcription for technical content

Benchmark report

30s english audio

solution	time
OiPer Desktop	1.5s
Lemonfox API	3.27s
Python Faster-Whisper	3.55s
OpenAI Whisper 1 API	6.46s

OiPer records the best speed while keeping transcription local and controllable.