| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| README.md | 2026-07-24 | 1.1 kB | |
| voxtype v0.1.3 source code.tar.gz | 2026-07-24 | 55.5 MB | |
| voxtype v0.1.3 source code.zip | 2026-07-24 | 55.8 MB | |
| Totals: 3 Items | 111.4 MB | 0 | |
voxtype 0.1.3
Whole-take dictation is now the default. On a parakeet engine in
toggle mode — the out-of-the-box setup on any Mac (Apple Silicon →
parakeet-mlx, Intel → parakeet) — plain voxtype now uses
segmentation = "hold":
- Press the hotkey, speak as long as you like, press again — the whole take is transcribed with full sentence context and typed in one go (maximum accuracy, a single edit to undo in most apps). No chopping at pauses.
- Prefer text to appear as you pause? Set
segmentation = "vad".
The default is smart: it only picks hold when the engine is
parakeet-family and mode is toggle, and falls back to vad for
moonshine or wake/vad modes — so it never produces an invalid combo. An
explicit choice in your config or a --segmentation flag is always
respected, and voxtype init leaves the key commented so the smart
default keeps working if you later change the engine or mode.
Upgrade:
:::bash
uv tool upgrade voxtype
macOS only (Intel and Apple Silicon). Everything runs on-device.