Saylo

Where your audio goes

The complete path of a recording, written to be quoted.

This page exists so you can copy it into a data-protection note, a donor report or a staff briefing. It lists every service that touches a file and what each one does. If something changes, this page changes first.

  1. 01

    Collect

    A file uploaded from the browser goes over TLS directly to our object storage; our application server issues the upload permission and does not relay the bytes. A voice note forwarded to the Telegram bot is fetched from Telegram’s servers by our bot and stored the same way.

    Saylo storage, Telegram
  2. 02

    Speech-to-text

    The audio is sent to ElevenLabs Scribe, which returns text, speaker labels and timestamps in the language spoken. ElevenLabs processes it under an API agreement; we do not opt in to any data-sharing or training programme.

    ElevenLabs Scribe v2
  3. 03

    Cleanup, summary, answers

    The transcript text, never the audio, is sent to a language model through our own gateway for punctuation cleanup, the reading version, speaker naming, the summary and any question you ask. It is not used to train anything.

    Google Gemini, via Saylo gateway
  4. 04

    Storage

    Audio, transcript, summary and chat history live in your account in our storage. They are not shared with other users, not indexed publicly and not used to train anything.

    Saylo storage
  5. 05

    Deletion

    Deleting a transcript removes the text, the summary and the source recording. Uploads that never became a transcription are purged automatically after 24 hours.

    You

Services that touch your data

Saylo (Deployka)Storage, application, billingOperator of the service. Holds your files, transcripts and account.
TelegramVoice-note delivery, if you use the botCarries the message you forward. Subject to Telegram’s own terms.
ElevenLabsSpeech-to-textReceives audio, returns text. No training opt-in.
Google Gemini via Saylo gatewayText cleanup, summary, answersReceives transcript text only. No audio.
CloudPaymentsCard paymentsReceives card details. Saylo never sees them.
ZitadelSign-inHolds your e-mail and password hash.

What we commit to

  • Your recordings and transcripts are never used to train any model.
  • Nobody on your team needs an account, an app or a browser extension.
  • Speakers are labelled per recording. No voice profile is kept across files.
  • Deleting from your account deletes from our storage, including the source recording.
  • This list of processors is complete. A new one is added here before it is used.

What we do not claim

  • No SOC 2 report or ISO 27001 certificate.
  • No HIPAA business associate agreement. Do not use Saylo for protected health information.
  • No on-premise or self-hosted option yet.
Where your audio goes · Saylo