Get started
Requirements
What Voxy needs to run, on macOS and on Windows.
Note Voxy for Windows is not out yet. This page describes both platforms so it is right the day it lands.
macOS
| Requirement | Detail |
|---|---|
| macOS | 14 Sonoma or later |
| Processor | Apple silicon or Intel — one disk image covers both |
| Memory | 8 GB. More helps if you turn on AI Refine |
| Disk | About 1 GB for the app and speech model. AI Refine needs roughly 5 GB more |
| Permissions | Microphone and Accessibility |
On Apple silicon the speech model runs on the Neural Engine. On Intel Macs it runs on the graphics card instead.
Windows
| Requirement | Detail |
|---|---|
| Windows | Windows 11, 64-bit — Intel/AMD and ARM |
| Processor | No graphics card needed. The speech model is fast on the processor alone; a GPU or NPU only makes it quicker |
| Memory | 8 GB. More helps if you turn on AI Refine |
| Disk | About 1 GB for the app and speech model. AI Refine needs roughly 5 GB more |
| Permissions | Microphone |
Internet
Dictation never needs the internet, on either platform. Voxy only goes online to:
- Download a speech model the first time you pick one.
- Check your licence.
- Look for a new version.
Everything else — recording, transcribing, refining, typing — happens on your own machine with no network calls at all.
Memory and AI Refine
AI Refine runs a small language model locally through Ollama, which is available for both macOS and Windows. Voxy will not let that model take more than a third of your machine's memory, and it steps aside automatically if the machine comes under memory pressure.
If you are tight on memory, the lighter model is a better fit — see AI Refine. You can also turn refining off completely and still dictate.
One licence, both platforms
Your licence covers macOS and Windows. Buying on one costs nothing extra on the other, and a device slot works the same way whichever you activate on. See License.