Privacy Policy

Seshari — last updated 18 September 2026

Data we collect

None. Seshari does not collect, transmit, sell or share any personal data. There is no telemetry, no crash reporting service, no advertising identifier and no third-party SDK that receives anything about you.

On-device processing

Transcription, dictation and text analysis all run entirely on your device, using Apple's on-device machine-learning frameworks. Your audio and your transcripts are never uploaded. There is no upload path in the app at all.

Network use

Seshari uses the network only to download, never to send:

• the speech-recognition model, once, on first run — after which transcription works fully offline;
• the optional generative and text-embedding models, only if you choose to enable those features.

These requests fetch public model files and contain no personal information about you. Purchases are handled by Apple's StoreKit, which communicates with Apple — not with us; we never see your payment details and receive no personal data from a purchase.

Microphone

Microphone access is used only while you are actively dictating, to turn your speech into text on your device. The recording is not sent anywhere. Whether an audio copy is kept on your device is your choice in Settings, and you can delete it at any time.

Files you open

When you drop in or share an audio or video file, Seshari reads it to transcribe it. It is not copied anywhere outside your device, and Seshari asks the system for access to nothing beyond the files you hand it and any folder you explicitly choose to watch.

Your data, and deleting it

Your transcripts live in Seshari's own storage on your device. Deleting a note removes its text and its audio copy; deleting the app removes everything. Since nothing was ever sent anywhere, there is nothing for us to delete on your behalf.

Windows

The Windows edition keeps the same rule — no upload path, nothing collected — with three differences that come from being sold outside an app store:

• Models. The speech-recognition model and the optional generative model are downloaded from huggingface.co, a public model host; the request fetches a public file and carries nothing about you. The small text-embedding model ships inside the app and is never downloaded.
• The price. When you open the Pro screen, or Settings, the app asks seshari.app/api/price for the current price so it can write it on the button. That is one plain web request — it carries no data about you beyond what any web request carries (your IP address, which Cloudflare, who hosts this site, sees in order to answer it) and nothing is stored from it. It is the only time the app contacts us.
• Buying Pro. The purchase happens on this site through Paddle, our merchant of record, which takes the payment, issues the invoice and handles sales tax; Paddle's own privacy policy governs the payment itself, and we never see your card details. From a purchase we receive your email address and the order id, and we keep that record so that we can re-send your key and honour refunds. Your Pro licence is a signed key sent to that address; the key encodes the address and the order id it was issued to, and the app verifies it on your PC — no account, no activation server, no check-ins. A refunded order is withdrawn by a list carried in the app's next update, not by any call home.

The app itself contains no analytics or telemetry of ours. It keeps a small diagnostic log on your PC that records what the app did — a window opened, a decode that failed — and never your text, your file names or your paths; it is not sent anywhere. Installation through the Microsoft Store, and whatever the Store records about it, is governed by Microsoft's privacy statement.

Children

Seshari collects no data from anyone, including children.

Changes

If this policy changes, the date at the top of this page changes with it.

Contact

Questions about privacy, or anything else: write to support@seshari.app.