This page explains the current shipped handling of practice-session data in Ducelis(霽序), without claiming broader policy machinery than exists today.
What session content is stored
We store the practice content needed to run and review a session, including user-entered text, generated counterpart replies, and the summaries and reports produced from that run. This data is used only to support rehearsal review and product operation, and is not repurposed for other aims.
Voice, transcripts, and generated reports
When voice is used, spoken input may be transcribed so the session can continue in text-backed form. Transcript artifacts and generated report content may be stored with the session so you can review what happened afterward. Voice data is not retained after processing is complete.
Data minimization guidance
Please avoid entering unnecessary personal data into practice sessions. When realistic detail is not essential, prefer fictionalized, masked, or anonymized examples. Names, events, and details mentioned in rehearsal scenarios should be replaced with non-real content.
Current product boundaries
These pages describe the current shipped product state. Some voice-forward behavior depends on current plan status, and Premium billing state is only reflected where the existing product already does so. We do not claim retention schedules, enterprise workflows, or other formal processes that are not defined here.
Demo avatar disclosure
Some demo avatars use AI-generated sample images for demonstration only. These images do not represent real people and should not be understood as portraits of any individual.
Privacy contact
If you have questions about privacy, please contact hello@ducelis.com.