Private preview terms.
Draft for a private adult US preview, dated October 14, 2026. Business identity and contact details must be added, and this draft must receive legal review before public launch.
What the service does
Mezmer provides AI-guided meditation and self-hypnosis for general wellness. It is not a licensed clinician or emergency service. You remain in control, may reject a suggestion, and may end a session at any time. Use it in a safe, stationary setting.
Required spoken interaction
A camera, microphone, and spoken intake are required. Camera images remain on your device. When you choose Speak, the app records up to thirty seconds and sends that check-in through Mezmer to ElevenLabs for transcription. Transcripts and coarse sensor measurements may also reach the configured script inference service to guide your session. Browser device permission alone does not constitute consent to every use; the session screen requests separate permissions.
Human script and transcript review
Authorized operators may inspect generated narration and spoken-check-in transcripts for safety and quality. They may approve, pause, replace, or end narration. These adjustments may happen without announcing each individual edit during the session. This is disclosed before you begin; we do not represent the service as exclusively automated or promise that no human can review it.
Optional listening permission
If you select the separate voice-replay choice, you authorize reviewers to replay your recent recorded check-ins during that session. This permission covers safety and quality review of those check-ins only. It does not authorize continuous ambient recording, sale of voice recordings, voice cloning, or future model training. Recordings are accessible in server memory for up to ten minutes and are cleared when the session ends. Speak alone and do not record other people without their consent.
Model comparisons
Sessions may be assigned to different configured models, and the assignment stays fixed within a session. We use limited session metadata, completion status, and voluntarily submitted feedback to compare quality. Participation does not imply that any experimental model is clinically proven. Operators may intervene when appropriate.
Data retention
Check-in transcripts and replay recordings are cleared from Mezmer session state when the session ends. Generated passages remain in memory for up to ten minutes after the session. Consent metadata, operator audit events, and model-comparison records are retained for fourteen days. User-requested pace preferences remain until deleted. Provider retention is governed by the relevant account configuration and provider terms; ordinary API access does not guarantee zero retention. Common generic narration audio may be saved for reuse. Personalized generated narration audio is not placed in the shared cache.
Training and voice ownership
You retain your rights in your contributions. The operational permissions above are limited to providing and reviewing the session. They do not grant permission to train future text or voice models. Any future training contribution will require a separate explanation and affirmative choice, including scope, retention, and withdrawal terms. No voice cloning permission is requested in this preview.
Age and availability
This preview is for adults eighteen or older in the United States. Service quality and availability can vary. These draft terms make no guarantee of habit change or treatment outcomes. Business ownership, support contact, privacy-request contact, and any applicable state-specific provisions must be completed before public availability.