Privacy Policy

Effective: August 17, 2026

Core promise

All speech recognition in Hotaru Voice runs entirely on your Mac. Your audio and recognized text are never sent to our servers. This is the core promise of the product. Any change to this policy will be announced in advance and require your explicit consent.

What the app sends

1. Required (minimum needed to provide the service)

• License validation (validity of your Pro license key)
• Free-plan usage (character counts as numbers only — never the text itself)

Dictation keeps working offline; these requests are made only when you are online.

2. Optional (off by default)

• Launch counts, recognition success/failure, error categories
• App version and macOS version

These quality statistics are sent only if you turn them on in Settings. They never include audio, text, or file names.

Anonymous ID

Communication uses a random UUID generated on your device. It is not derived from hardware or personal information, and you can reset it anytime in Settings.

When you use cloud features

Only when you explicitly invoke a cloud feature (instruct mode, voice chat) is the target text sent to the processing server. It is not used for training. If you do not use these features, no text ever leaves your device.

Training data (opt-in)

Saving audio/correction pairs for accuracy improvement is off by default and, when enabled, stays on your device. It is shared with us only by your own action.

Where data goes and how long it is kept

Data is sent to our server on Cloudflare (api.hotaruvoice.com). Usage and statistics are retained for up to 24 months. Website access logs follow Cloudflare's standard operation.

Contact

Questions? Reach us via GitHub Issues.

Changes

Revisions will be announced on this page. Changes that add new data collection will not apply without explicit re-consent in the app.