Privacy & recording policy

No audio is captured or stored until a parent or guardian explicitly allows voice recording for their child on the parent dashboard. Consent can be withdrawn at any time, which immediately stops new recordings from being saved.

• Account data: email address and display name for parents and clinicians.
• Child profile: first/last name, age, avatar, and the speech sounds being targeted.
• Practice attempts: the target word, the AI transcription of what was said, whether it matched, and a timestamp.
• Short voice clips (up to 5 seconds each) stored in a private bucket that is not publicly readable.

Only the child's linked parent or guardian and their assigned clinician can play back recordings. Playback uses links that expire after five minutes and are generated on request — recordings are never served from a public URL.

Each clip is sent to a speech-to-text model solely to produce a transcription used to score the attempt. Clips are not used to train models, are not sold, and are not shared with advertisers or any third party beyond the transcription provider.

Storage per child is capped, and older clips are pruned automatically as new ones come in, so we keep only a recent review window rather than an indefinite archive. Transcriptions and accuracy history are retained for progress tracking.

A parent or guardian can request deletion of their child's recordings, transcriptions, and profile at any time by contacting their clinician or writing to us; we remove the data from active systems on receipt.

TinyTalk AI supports home practice between sessions. It does not diagnose, and AI scoring is a practice aid — not a substitute for a qualified clinician's judgment.