Privacy First

How Vocalinux handles your voice

Vocalinux is built for privacy. Local engines process speech on your machine. Remote API is optional and only talks to a server you set.

Local engines

Local engines do not need the internet after the model is downloaded. Processing stays on your device unless you turn on Remote API.

No Cloud Processing

Local engines do not send your voice to our servers. Remote API sends audio only to the URL you configure.

Open Source

Fully open source under AGPL-3.0. Inspect the code yourself on GitHub. No hidden telemetry or data collection.

Data Collection

We do not collect any personal data. This includes:

  • No voice recordings or audio data
  • No transcribed text
  • No usage analytics or telemetry
  • No device fingerprints
  • No IP addresses or location data

How We Process Your Voice

When you dictate, here's exactly what happens:

  1. 1. Audio Capture - Your microphone records audio temporarily in RAM
  2. 2. Local Processing - Audio is processed by whisper.cpp, Whisper, or VOSK on your CPU/GPU
  3. 3. Text Output - Transcribed text is injected into your active application
  4. 4. Immediate Deletion - Audio data is discarded immediately after processing

Note: Local engines do not write audio to disk or send it over the network. Remote API sends audio to the server you configure.

Configuration Data

Vocalinux stores configuration locally at ~/.config/vocalinux/:

  • Settings like preferred engine, model size, and language
  • Custom keyboard shortcuts
  • Window and tray preferences

This data is stored in plain text YAML format. You can view, edit, or delete it at any time.

Vocalinux vs Cloud Dictation Services

FeatureVocalinuxCloud Services
Data leaves deviceOnly with Remote APIAlways
Internet requiredOnly for Remote API or a model downloadYes
Source code availableYes (AGPL-3.0)No
Voice recordings storedNoVaries by service
CostFree foreverSubscription

Questions About Privacy?

If you have any questions about our privacy practices or want to verify our claims, please reach out:

Open an issue on GitHub