We don't have your data.
This is the short version of a privacy policy, because there isn't much to say. Scream runs open-source speech models entirely on your device. Nothing you say or type is sent to us or anyone else.
How Scream works
Scream uses open-source speech recognition models that are downloaded once and then run 100% locally on your device. Recording, transcription, cleanup and typing all happen on your own hardware. Scream works with your internet switched off — because it doesn't need it.
What we collect
Nothing. Specifically, we do not collect, store, or have access to:
- Your voice recordings or audio
- The text Scream produces
- The apps you use or what you type into them
- Your name, email, or any account details — there are no accounts
Audio is processed in memory to produce text and is discarded afterwards. We don't run servers that receive it, so there's nowhere for it to go.
Open-source models
The speech models Scream uses are open source, meaning anyone can inspect how they work. They run on your machine like any other local software — no remote inference, no "improving our models with your data".
Contact form
If you choose to write to us through the contact page, the details you enter are used only to reply to you. They are not shared or sold.
This website
This site loads fonts from Google Fonts. The "Try your mic" demo on the home page analyses your microphone level inside your browser only; nothing is recorded or uploaded.
Changes
If anything about this ever changes, we'll update this page. Questions? Get in touch.
Last updated: October 22, 2026