FAQ

If your question is not here, the troubleshooting guide covers specific symptoms, and privacy architecture answers data questions in far more depth. Anything still unanswered: shimoverse@gmail.com.

Frequently asked

Is OpenVoiceFlow really free?

Yes. It is MIT-licensed, has no paid tier, no account, and no trial period. Your Mac provides the compute, so there is nothing to charge you for. If you enable optional cloud cleanup you pay your chosen AI provider directly for tokens — typically cents per month — and nothing to us.

Does my voice leave my Mac?

No. Audio is transcribed on-device by Whisper and discarded once text exists. There is no code path that uploads audio. If you turn on cloud AI cleanup, the resulting text is sent to the provider you chose under your own key; choosing Ollama or leaving cleanup off keeps everything local.

Does it work offline?

Yes. After the one-time model download, dictation works in airplane mode. With cleanup off, which is the default, no network request is made at any point.

Which Macs are supported?

macOS 14 Sonoma or newer, on both Apple Silicon and Intel. One universal build covers both. For macOS 12 to 13 a retained 0.3.6 Apple Silicon build is available, but it is end-of-life.

Is there a Windows, Linux, iPhone, or Android version?

Not today. OpenVoiceFlow is macOS-only. iOS and Android versions are in development but cannot be downloaded yet.

How accurate is it?

That depends on your microphone, accent, vocabulary, and model choice far more than on anything we could benchmark. Whisper's larger models are widely regarded as excellent. The honest test takes a minute: dictate a real paragraph of your own work and judge the result.

Can I say punctuation out loud?

Not in version 0.5.5 — saying 'comma' produces the word 'comma'. Whisper punctuates from sentence structure automatically, and optional AI cleanup fixes grammar and removes filler words if you want more.

Why is my dictionary not working?

The personal dictionary, Know-Me profile, and per-app styles are all applied during AI cleanup. With cleanup set to None, which is the default, they have no effect. Turn cleanup on, or use local Ollama if you want that behaviour without sending text to a cloud.

Which languages does it support?

Eleven in the picker: English, Spanish, French, German, Italian, Portuguese, Dutch, Hindi, Japanese, Chinese, and Korean. Accuracy outside English depends heavily on model size, so use the largest model your Mac handles comfortably.

What happens if the AI cleanup service fails?

You get your raw transcript. Cleanup is designed to fail open: on a missing key, an error, a timeout, or an unparseable response, the app inserts the on-device transcription rather than losing your words.

Does it collect any analytics?

The app contains no telemetry whatsoever — no usage pings, no crash reporting, no feature counters. The website uses privacy-friendly anonymous page-view analytics, which is entirely separate from the app.

How do I get support?

Email shimoverse@gmail.com. Include your macOS version, your Mac's chip, your OpenVoiceFlow version, and what you saw versus what you expected. Because the app has no telemetry, your description is all we have to work from.

Can I use OpenVoiceFlow at work with confidential material?

The architecture is designed for exactly that case: audio never leaves the machine, and with cleanup off or pointed at local Ollama, neither does text. Your organization's own policy still applies, and the source being open means your security team can verify the claims rather than trust them.

Can I use the code in my own project?

Yes — it is MIT-licensed and free for any use. Keep the credit visible, and we would genuinely like to hear about it at shimoverse@gmail.com.