Vox
Hebrew audio in, synced transcript out.
Local Hebrew audio transcription for macOS. Vox runs the ivrit-ai whisper-large-v3 model entirely on-device — no upload, no account, no API key — and produces a self-contained HTML page: the full transcript synced to a built-in audio player, click any line to jump there.
macOS 14 · Apple Silicon · Right-click → Open on first launch
What it does
Fully on-device
Runs the ivrit-ai whisper-large-v3 model entirely on your Mac. No upload, no account, no API key — nothing leaves the machine.
Any audio or video
Drag in mp3, m4a, wav, mp4, mov — Vox converts it with ffmpeg and transcribes with whisper.cpp.
Synced transcript page
Output is a self-contained HTML page — transcript synced to an audio player, click a line to seek, with follow-along highlighting.
Auto chapters
Optionally, a local LLM (Gemma 3 via Ollama) splits the transcript into categorized chapters with Hebrew titles.
Build it yourself
Vox's full source is on GitHub. The Xcode project is generated with
XcodeGen — install it, run ./build.sh, or open the
project directly. It also drives whisper-cpp and
ffmpeg (brew install whisper-cpp ffmpeg).
git clone https://github.com/MotionPeak/vox.git