Speak. Your words appear at the cursor.
I’m building VOCO to make dictation on Linux fast, simple and private. Speech recognition runs on your computer. No account, subscription or cloud transcription.
- Live dictation: words and punctuation appear as you speak.
- One shortcut: press
Alt+Dto start, then again to stop. - Local English model: NVIDIA Nemotron runs on the CPU; no GPU needed.
- Tray-first: VOCO stays out of the way while you work.
Download the Debian package and checksum file from Releases. In the download folder:
sha256sum -c voco_latest_checksums.txt
sudo apt install ./voco_latest_amd64.debRun the install command only if checksum verification passes. Ubuntu x86_64 is the reference platform. See the installation guide for setup, upgrades, source builds and Linux compatibility. The newest development candidate may be newer than the latest public download; release status keeps that distinction clear.
- Open VOCO and check your microphone.
- Focus a text field, press
Alt+D, and speak. - Press
Alt+Dagain to finish. Review your text before sending it.
VOCO delivers text through clipboard paste and never presses Enter. Keep the same field focused. Some protected or custom editors may not accept delivery; interrupted transcripts stay available in VOCO for recovery. Usage and recovery.
Audio stays on your device. Optional local performance logs contain timings and counts, not recordings or dictated text. Update checks contact GitHub. Diagnostics · Security.
Curious how it works? Explore the visual code guide.
Start with development setup, the code map and AGENTS.md. Tests and benchmark methods are indexed in testing.
VOCO is MIT licensed. Bundled models and native libraries have their own licenses.