Privacy policy

Effective November 11, 2026. Covers Whisper Unchained Server for Windows, Whisper Unchained for Android (both the Google Play version and the Full version from GitHub), and this website.

Who we are

Whisper Unchained is free, open-source software (MIT License) made by Dominic Shirazi at Cyrus Studio, Oregon, USA. Questions about this policy go to Dominic@cyrus-studio.com. Because the source code is public, anyone can check what the apps do: Windows server, Android app.

Where your voice and text go

You choose one of these in Settings, and you can change it at any time:

• On this PC. Speech recognition runs on your own computer. Audio and text stay on it.
• On another PC of yours. Audio travels from your phone or laptop to your own PC over your home network, or over Tailscale if you set it up. It goes nowhere else.
• A cloud service. If you choose Groq, OpenAI or another OpenAI-compatible service, the audio is sent to that service using the API key you gave the app. What happens to it there is covered by that service's own privacy policy, not this one. Pick the local options if you don't want that.

The optional AI editing features ("prompt AI", cleanup and mean mode) send the transcribed text to the AI server you set up for them. Usually that is software on your own PC, such as Ollama or LM Studio. If you point it at a cloud AI service instead, that service's privacy policy applies.

Whisper Unchained for Android

Permissions and why the app asks for them

• Microphone: to record what you say, only while you are recording. You start and stop every recording yourself.
• Display over other apps: to show the floating microphone button on top of the app you are typing in.
• Accessibility service (optional): see below.
• Notifications: Android requires a notification while the microphone button is active.
• Run at startup (optional): to bring the floating button back after your phone restarts, if you turn that on.
• Vibration: small haptic ticks when you adjust the style while recording.
• Internet: to send your recording to the server you chose, and to receive the text back.
• No camera permission. When you scan the pairing QR code shown by your PC, Google Play services' code scanner handles the camera on your phone, and the app receives only the text inside the code (your server's address and key), never the image.

Accessibility service

Whisper Unchained uses Android's Accessibility service for one job: to notice when a text box is selected, so the floating button can appear, and to paste your transcribed text into that box. It does not read, record or store what is on your screen or what you type, and nothing it sees is sent anywhere. It is optional. If you leave it off, the app copies your text to the clipboard instead and you paste it with one tap.

Audio from your phone (Full version only)

The Full version, downloaded from GitHub and not from Google Play, can also transcribe sound playing on your phone, such as a video or podcast. Android asks your permission each time, and shows that capture is on. Phone calls are excluded by Android. The Google Play version doesn't include this feature.

What stays on your phone

• Your settings, your vocabulary and your server address are kept in the app's private storage. Any API key is encrypted with the Android Keystore.
• The latest recording is kept in the app's cache until the next recording replaces it. Uninstalling the app removes it, along with everything above.
• Android backup is turned off for the app, so none of this is copied to Google's cloud backup.

Whisper Unchained Server for Windows

• Your settings, vocabulary and name corrections are kept in files in the app's folder on your PC.
• Transcript logging is off by default. If you turn it on (Settings > Audio Corpus), text is saved only on your PC, and the same screen deletes it.
• Keeping audio is off by default. If you turn it on, recent recordings are kept on your PC for testing models. If you say one of the "tag this clip" phrases, that recording is kept on your PC permanently. None of it is uploaded.
• Other devices can use your server only over your own network or Tailscale. You can require an API key for them in Settings > Phone.
• When you uninstall, Windows asks whether to delete your settings, transcripts and models.

Downloads during setup

To install, the setup downloads the tools it needs from their usual public sources: Python and its packages (through uv and the Python Package Index), NVIDIA's libraries if you have an NVIDIA card, the speech recognition model from Hugging Face, and the service manager NSSM. If you choose to install Tailscale, it comes from tailscale.com. Like any download, those sites see your internet address when you fetch them. Nothing about you or your use of the app is sent with these downloads.

This website

This website has no cookies, no analytics and no tracking scripts. It is hosted on Cloudflare, which, like any web host, handles the standard technical information your browser sends (such as your internet address) to deliver the page. The download buttons take you to GitHub, where GitHub's privacy statement applies.

Sharing, selling and advertising

We don't sell, share or rent any personal information, because we don't have any. There are no ads in either app.

Children

Whisper Unchained is a general-purpose tool and isn't directed at children under 13. We don't knowingly collect information from anyone, including children.

Deleting your data

Everything the apps keep is on your own devices, so you are in control of it: delete it from Settings, or uninstall the app. We have nothing on our side to delete. If you used a cloud service, contact that service about anything it holds.

Changes to this policy

If the apps ever start handling data differently, we will update this page and the date at the top before that version is released.

Contact

Dominic Shirazi, Cyrus Studio: Dominic@cyrus-studio.com. You can also open an issue on GitHub.