Native voice transcription for macOS and iOS. Use private on-device transcription with Nemotron or Apple SpeechAnalyzer, or bring your own Groq/OpenAI API key.
Settings — Dictation · AI Polish · General · About
Menu bar
- Hold
fn→ speak → release to paste text anywhere - Hold
fn + shiftto apply AI polish - Private on-device transcription with Nemotron 3.5 ASR
- Text is pasted after you release
fn - Apple SpeechAnalyzer support on macOS 26+
- Cloud transcription with Groq, OpenAI, or a custom OpenAI-compatible endpoint
- Menu bar app, no telemetry
- SSH terminal with voice dictation
- Floating dictation controls
- On-device Nemotron transcription with live preview
- Cloud transcription with Groq, OpenAI, or custom endpoints
- Saved SSH connections with Keychain password storage
- Hack Nerd Font included for terminal prompts
-
Download the latest release from Releases.
-
Extract
Omri-vX.X.X-apple-silicon.zip. -
Remove quarantine:
xattr -rd com.apple.quarantine Omri.app
-
Move
Omri.appto Applications. -
Launch Omri and grant microphone access.
-
Pick a transcription provider in Settings.
Hold ⌥Space to dictate. The hotkey needs no extra permission.
- Nemotron (On-Device): private, offline after model download, ~40 languages, no API key
- Apple (On-Device): private native SpeechAnalyzer on macOS 26+
- Groq/OpenAI: cloud transcription with your own API key
- Custom: OpenAI-compatible transcription endpoint
- macOS 26.0+
- Apple Silicon recommended for on-device Nemotron
- Microphone permission (required)
- Accessibility permission (optional) — auto-inserts the transcript into the active app; without it, the text is copied to the clipboard
- Input Monitoring (optional) — hold the fn key to dictate; the ⌥Space hotkey works without it
- iOS 26.0+
- Microphone permission
- Internet only for cloud APIs or first on-device model download
git clone https://github.com/nasedkinpv/omri.git
cd omri
open Omri.xcodeprojBuild schemes:
xcodebuild -project Omri.xcodeproj -scheme Omri -configuration Debug build
xcodebuild -project Omri.xcodeproj -scheme OmriiOS -sdk iphonesimulator build- On-device audio never leaves your device.
- Cloud APIs are used only when you select a cloud provider.
- API keys and SSH passwords are stored in Keychain.
- No telemetry or tracking.
- Temporary audio files are deleted after processing.
- Accessibility permission (macOS) is used only to insert transcribed text at your cursor in the app you're using. Omri never reads screen contents; if declined, it falls back to copy and paste.
MIT License — see LICENSE.
On-device transcription uses the NVIDIA Nemotron 3.5 ASR model, Copyright © NVIDIA Corporation, licensed under OpenMDW-1.1. CoreML conversion and runtime by FluidAudio (Apache-2.0).
Created by Ben Nasedkin.