Third-party components and licenses
Applies to WhisprDesk 1.7.4 · Updated 2026-10-25
WhisprDesk converts speech to text only. It does not include or use any text-to-speech component. It is built on the open-source components listed below, each used under its own license.
Components
| Component | What it does in WhisprDesk | License |
|---|---|---|
| OpenWhispr | WhisprDesk is derived from the OpenWhispr project | MIT |
| whisper.cpp | Runs OpenAI Whisper models on your Mac for local transcription | MIT |
| OpenAI Whisper models | Local speech-to-text models, downloaded to your Mac only when you choose one | MIT |
| NVIDIA Parakeet TDT 0.6B v3 | Local speech-to-text model by NVIDIA, downloaded to your Mac only when you choose it | CC-BY-4.0 |
| sherpa-onnx | Runs the Parakeet model on your Mac | Apache-2.0 |
| ONNX Runtime | Inference engine used by sherpa-onnx | MIT |
| FFmpeg | Converts audio to the format the speech models need. WhisprDesk uses an audio-only build that we compile ourselves. The exact FFmpeg source code and our build script are published with each WhisprDesk release, next to the download, at github.com/Whisprdesk/releases. | LGPL-2.1-or-later |
| llama.cpp | Runs a local AI cleanup model on your Mac, only if you choose one | MIT |
| Electron and Chromium | The app framework | MIT, with BSD-style and other permissive licenses for Chromium components |
| Noto Sans | The typeface used in the app | SIL Open Font License 1.1 |
About NVIDIA Parakeet
Parakeet TDT 0.6B v3 is a speech-to-text model created by NVIDIA and released under the Creative Commons Attribution 4.0 license (CC-BY-4.0). WhisprDesk downloads a version converted for sherpa-onnx by the k2-fsa project, and only when you choose to download it. Parakeet turns speech into text. It does not generate speech.
License notices
The full license texts and copyright notices for these components are included with the app. If you would like a copy, email support@whisprdesk.com.