Your voice never leaves your servers

Speech is transcribed and translated on the fly, in memory, then not retained. No audio file is recorded or stored unless you explicitly enable an export.

Speaker identification “who speaks, when” works in real time, for the duration of the session. By default, no voice print is kept after the session.

If recurring speaker recognition is enabled (optional), the voice print is biometric data (GDPR art. 9): it requires explicit consent, is encrypted at rest, isolated per organization, and erasable in one click. It is never a replayable voice.

Synthetic voices and any voice cloning (from a sample provided with consent) run locally. Models and voices stay on your infrastructure.

Your organization remains the data controller. TranscriLive / Ambassadia acts as a technical processor, under a Data Processing Agreement (DPA, GDPR art. 28) defining purposes, durations and security measures.

Access, rectification, erasure, restriction, portability, objection. As processing is local and without audio retention by default, exercising these rights is immediate and under your control.

Only the data strictly necessary for transcription, translation and accessibility is processed, for the sole duration of the session.

None. As processing is on-premise, your data never passes through any third-party service or server outside your infrastructure.