Theme Text contrast Text size Window mode Layout

Settings are saved locally in your browser.

Browser mic device

System audio device is configured via --device at server startup.

Engine for "Speak context" mic button

Whisper runs on your server — better for technical terms. Browser uses built-in speech recognition.

Extraction interval

How often new transcript is sent to Claude for concept extraction.

Overlap segments

Segments carried over between extraction windows to catch boundary topics.

Card visibility

Seconds an Incoming card stays visible before auto-expiring to History.

Transcription engine

faster-whisper uses CTranslate2 (CUDA support). whisper.cpp uses GGML (cross-platform). Changing engine resets model to Base.

Transcription model Tiny (fastest, least accurate) Base (default) Small (better for technical terms) Medium (slower, more accurate) Large v3 (slowest, most accurate) Language Auto-detect English Spanish French German Italian Portuguese Dutch Polish Russian Chinese Japanese Korean Arabic Hindi Turkish Swedish Danish Norwegian Finnish Czech Romanian Hungarian Greek Hebrew Thai Indonesian Malay Vietnamese Ukrainian

Setting an explicit language improves accuracy, especially with accented speech. Use "Auto-detect" for multilingual meetings.

Help improve Ori by sharing anonymous usage counts and timing data. Analytics never includes meeting content, audio, transcripts, or extracted terms.