Settle in. We’ll find your pace together.
Nothing to get right.Just a place to begin.
Camera & spoken check-in required · human review disclosed
Following your chosen pace
Camera images stay on your device. Spoken check-ins are recorded when you choose Speak and sent through Mezmer to ElevenLabs for transcription.
Authorized human reviewers can read your check-in transcripts and generated scripts, pause narration, replace passages, or adjust sound layers you have enabled for safety and quality. Sessions may use different AI models for comparison. Each edit need not be announced during playback.
Opted-in recordings stay in server memory for up to ten minutes and are cleared when the session ends. This does not authorize model training or voice cloning.
Checking the voice connection…
Escape ends playback and releases both devices.
Say how your evening feels and what you’d like help with.
Recording starts only when you choose Speak.
Maximum 30 seconds. Speak alone so other people aren’t recorded. Your recording is sent to ElevenLabs for transcription.
Camera is off
Vocal analysis uses quiet listening windows.
Measurements use a rolling window. Pitch appears only when a voiced sound is detected and narration is quiet. No emotion or trance score is inferred.
Camera images remain on your device. Acoustic measurements are analyzed locally. Spoken check-ins are recorded only through the Speak control and uploaded for transcription.
Authorized reviewers can inspect scripts and check-in transcripts while your session is active, and adjust the narration. Voice replay requires the separate optional checkbox.
Check-in transcripts and replay recordings are cleared when the session ends; recordings also expire after ten minutes. Generated passages remain in memory for up to ten minutes after the session. Consent metadata, review audit events, and model-comparison feedback are retained for fourteen days. Preferences are saved only when requested.
The voice connection is being checked.
These permissions do not authorize training or voice cloning. Model-training contributions will need a separate choice.
Presets are synthesized sound textures. Your own file stays in browser memory; up to 20 MB and five minutes. Start with a low device volume.
An optional repeated sound you can associate with settling. Its effect is not established; you can turn it off at any time.
Optional tones, not hidden suggestions. Hearing varies, and this device may not reproduce very low or very high frequencies. These controls do not measure your hearing threshold or sound pressure. Do not turn up the device to hear a missing tone.
Tone levels have a quiet digital cap. This is not a guarantee of acoustic safety or an anchoring effect. Reviewers can adjust layers you have enabled; they cannot enable them for you.
Sound begins after your spoken check-in. All layers pause while you speak.
Your rating supports the disclosed model comparison and is retained for fourteen days. It is not a model-training contribution.