MyWare — desktop software

Your voice stays on your machine.

MyWare makes offline tools for speech. Scribe turns recordings into speaker-labelled transcripts; Voice types what you dictate into whatever window you're working in. Both run the speech models locally — no accounts, no cloud, no telemetry, and your audio never leaves the device.

The differentiator

Privacy you don't have to configure.

This isn't a settings page promise. Both apps are built so there is nothing to turn off and nothing to trust. The one network call they make is downloading the speech models — once, with checksums.

No accounts
A licence is a key, not a login. Nothing to sign into, nothing to leak.
No telemetry
No analytics, no crash beacons, no usage phoning home — in the apps or on this site.
No cloud
Speech models run on your CPU or GPU. Audio and transcripts stay on the device.
Offline by default
After the model download, every feature works with the network cable pulled.

How it's built

Native software, end to end.

  • One binaryBoth apps are native Rust executables — a real desktop GUI and a command-line mode. No Electron shell, no Python runtime to manage, no background services you didn't ask for.
  • Named enginesScribe diarizes with sherpa-onnx (pyannote segmentation + wespeaker embeddings) and transcribes with whisper.cpp. Voice recognises speech with NVIDIA Parakeet TDT 0.6B v3 via sherpa-onnx. The models and their licences are disclosed, not hidden.
  • Light & darkBoth apps ship light and dark palettes and follow your desktop by default — this site does too.

Start free. Stay when it's useful.

Both apps have a working free tier — the full transcription pipeline in Scribe, the whole dictation loop in Voice. Pro is a one-time purchase that adds the heavier models and the extras, not a subscription that holds your work.