VoiceInk for Windows

No. VoiceInk by Beingpax has no Windows desktop build. Its desktop app requires Apple Silicon and macOS 14.4 or later, and the developer also offers a separate iOS app. The macOS source is available under GPLv3, but the project does not target Windows or Linux.

If you want VoiceInk's core idea (local push-to-talk dictation with a one-time price) on a Windows PC, Whisperstream is the closest match. It runs the same NVIDIA Parakeet model family that VoiceInk offers locally, except natively on Windows and on your CPU.

VoiceInk and Whisperstream both offer a one-time price, local transcription, and a fast push-to-talk loop. VoiceInk 2.0 also reaches beyond basic dictation with configurable Modes, optional cloud providers, an assistant, file queues, and history analysis. Here is the practical shape of each, side by side.

Pricing

$25 lifetime (Solo tier, single Mac); or build from source under GPLv3

Platform

macOS 14.4+ (Apple Silicon only)

On-device

Yes (Whisper via whisper.cpp, plus NVIDIA Parakeet via FluidAudio)

Best for

Mac dictation, files, configurable Modes, and an indie ethos

• $25 lifetime for a single Mac (Solo tier), with $39 and $49 tiers for more devices, or build from source under GPLv3 for free.
• Runs Whisper and NVIDIA Parakeet locally on your Mac via whisper.cpp and FluidAudio. Parakeet V3 is also available in Whisperstream on Windows.
• VoiceInk 2.0 adds Modes for app- and task-specific workflows, VoiceInk Assistant, broader model support, and productivity analysis.
• Live dictation, real-time transcription, multi-file audio and video transcription, and searchable history with playback and export.
• Local transcription is available, with optional cloud transcription and AI enhancement providers when you enable them. The source is on GitHub under GPLv3.

VoiceInk is the indie, open-source pick. Version 2.0 combines local Whisper and Parakeet transcription with configurable Modes, real-time models, VoiceInk Assistant, multi-file transcription, and searchable history. The Solo tier is a one-time $25 for a single Mac, and the source is on GitHub under GPLv3.

Browse the VoiceInk Wall of Love and the same notes repeat: a one-time payment, a clean native interface, and no subscription. Made by an indie developer who goes by Pax, it has become a favorite for Mac users who want a local-first dictation app with optional AI workflows.

Both apps offer NVIDIA Parakeet V3 as a local transcription option. VoiceInk uses FluidAudio on Apple hardware; Whisperstream runs an ONNX version on Windows CPU and adds Qwen3 ASR for more languages. Sharing a model family does not guarantee identical output because each app has its own audio, decoding, and formatting pipeline.

The useful takeaway is model availability, not an app-to-app accuracy promise. With local transcription selected in either app, audio stays on the machine; VoiceInk also offers optional cloud providers.

Pricing

$29 once

Platform

Windows 10/11 x64

On-device

Yes (NVIDIA Parakeet and Qwen3 ASR via ONNX Runtime, on CPU)

Best for

Live push-to-talk on Windows

• $29 one-time and activates on up to 2 of your own PCs, with a 30-day refund and a 7-day free trial with unlimited dictation.
• NVIDIA Parakeet and Qwen3 ASR via ONNX Runtime, running entirely on your CPU. Audio never leaves your PC.
• Push-to-talk hotkey (default right shift) that delivers text to many compatible focused fields.
• An optional searchable transcript history with optional audio playback, encrypted at rest on your device and never synced to a server.
• Optional AI cleanup can use a downloadable local model on a capable GPU, with no API key, or a provider you configure. It is off by default.
• A Windows-first, one-time-purchase alternative with local transcription and no required cloud account.

Whisperstream is built for Windows the way VoiceInk is built for the Mac: a one-time purchase, audio that never leaves your machine, and a fast push-to-talk loop you can map to whatever key feels right. The engines are NVIDIA Parakeet and Qwen3 ASR, running on your CPU through ONNX Runtime, so there is no GPU requirement and no cloud round-trip.