Offline

Offline Voice Dictation

Local engines keep speech recognition on your computer using whisper.cpp, VOSK, or OpenAI Whisper models on your hardware. Remote API is optional and talks only to a server you configure.

Works without a network

After the model is downloaded, local engines work with no internet. Remote API needs the server you configured.

Zero Data Collection

No telemetry, no analytics, no usage tracking. Local engines have nowhere to send audio unless you enable Remote API.

No Account Required

No signup, no login, no password to forget. Download and run - that's it.

Your Voice Stays Yours

Voice data is biometric. Your voice patterns, accent, and speech habits are yours alone.

Offline vs Cloud Dictation

FeatureVocalinux (Offline)Cloud Dictation
Requires internet for local dictation
Voice data uploaded to vendor servers
Account/signup required
Subscription cost
Works in airplane mode after model download
Vendor can shut down service

Local engines

Local engines keep audio on the machine

whisper.cpp, Whisper, and VOSK process speech on your hardware. That is a default for those engines, not a legal or compliance promise.

Works after the model is downloaded

Once the model is on disk, local engines keep working without a network. Remote API still needs the server you set.

Remote API is opt-in

If you point Vocalinux at a server you configure, audio goes there. Local engines do not.

How Offline Processing Works

Vocalinux uses speech recognition models that run entirely on your CPU or GPU:

  • whisper.cpp - C++ port of OpenAI Whisper, runs locally with GPU acceleration
  • VOSK - Lightweight offline engine, minimal resources
  • OpenAI Whisper - PyTorch-based, runs on NVIDIA GPU or CPU

Models are downloaded once (74MB-3.0GB depending on size). After that, local engines process on the machine. Remote API is the exception: it sends audio to the server you configure.

Verify It Yourself

Don't trust - verify. Vocalinux is open source. You can:

  • Read every line of code on GitHub
  • Run with --debug and watch network activity. Local engines should stay quiet during dictation.
  • Monitor with tcpdump or wireshark. Local engines should stay quiet during dictation.
  • Disconnect from the internet and check that a local engine still dictates

Truly Private Voice Dictation

Install Vocalinux and use a local engine if you want speech recognition on the machine. Remote API stays optional.