Voicy is derived from
SuperDictate (forked at v0.2.40,
commit 83dd7e4), which is itself derived from
Parakey, originally created by
Richard Courtman. Both are distributed under the MIT License and both carry
the copyright notice Copyright (c) 2026 Richard Courtman, which is preserved
in LICENSE alongside the copyright on Voicy's own changes.
SuperDictate contributed substantial product, interface, service-management, history, statistics, and dictation-workflow modifications on top of Parakey. Voicy adds its own renaming, modularisation, interface, and local-statistics work on top of SuperDictate, and removed the self-update mechanism, the menu-bar item and the public release tooling that came with the fork.
Voicy links FluidAudio
(Apache License 2.0) statically, so a distributed Voicy.app contains its
code. The required license texts are included in third_party/:
| File | Covers |
|---|---|
third_party/FluidAudio-LICENSE.txt |
FluidAudio, Apache License 2.0 |
third_party/fastcluster-LICENSE.md |
fastcluster (BSD-2-Clause), vendored by FluidAudio |
third_party/vbx-LICENSE.md |
VBx (Apache License 2.0), vendored by FluidAudio |
FluidAudio is the only package dependency; see swift/Package.resolved for
the pinned revision.
The speech model is not part of this repository and is not redistributed here.
FluidAudio downloads it on first launch from Hugging Face
(FluidInference/parakeet-tdt-0.6b-v3-coreml, Apache License 2.0), a Core ML
conversion of NVIDIA's
nvidia/parakeet-tdt-0.6b-v3.
The model weights remain under their own upstream license; check the model
card before using Voicy's output in a context where that matters.