Privacy
Privacy policy.
Short version: Celesa does not collect your voice, your text or your identity. Last updated 31 October 2026.
The Celesa app
- Your voice is recorded into memory while you hold the key, turned into text on your Mac, and then discarded. Audio is never saved or sent.
- Your text is tidied on your Mac and pasted where your cursor is. It is never sent to us or anyone else.
- History. Your last 50 dictations are kept as text in a file in your Mac’s Application Support folder, so you can find them again. You can delete that file at any time.
- Settings, dictionary and snippets are stored in the same folder on your Mac.
- No account, no analytics, no tracking. The app does not identify you.
When the app connects to the internet
- Model download. On first launch, Celesa downloads its speech and writing models from Hugging Face (huggingface.co). Hugging Face sees a normal download request, including your IP address.
- Update check. Celesa checks a file on GitHub (raw.githubusercontent.com) for new versions, and downloads updates from GitHub when you choose to install one. GitHub sees a normal request, including your IP address.
Neither request contains anything you have said or written.
This website
celesa.ai uses no cookies, no analytics and no third-party scripts. It is served by Cloudflare, which processes standard request data such as IP addresses to deliver and protect the site.
Contact
Questions about privacy: hello@celesa.ai.