Research-preview data agreement.
The three required purposes presented before guided enrollment.
Who is asking for your consent
Silent Speech is a research preview from interfaces.inc. The research preview and service are operated by Dafdef, inc., doing business as Interfaces. Your consent is given to Dafdef, inc.
The research preview requires agreement to all three purposes below. If you do not agree to every purpose, do not participate.
1. Personalization and evaluation
You agree to provide an initial 150 prompted front-camera recordings and, if you affirmatively choose the optional booster, 50 additional prompted recordings. This includes prompt text, word-pacing timestamps, capture-quality metadata, corrections you explicitly submit, and derived model data. We process this data to create, evaluate, operate, debug, and retrain a visual-speech model for you.
2. Model research and training
You agree that your prompted enrollment recordings, labels, timing, quality metadata, corrections, and derived features may be used to develop, train, evaluate, and improve interfaces.inc AI models, including current and future speech, interface, and foundation models, and to measure performance across participants. We do not use this data to identify you, and we do not sell it.
3. Product-use contribution
You agree that every completed Interfaces for Mac dictation may contribute its synchronized front-camera video, microphone audio, local transcript and timing, capture-quality metadata, and limited device metadata to develop, train, evaluate, and improve the models covered by this agreement. You also agree that when you explicitly submit a correction to a decoded silent-speech message, the associated clip, decoded text, corrected text, timing, quality metadata, and derived features may be retained and used to improve and evaluate Silent Speech models. Ordinary silent-speech product-use clips without a submitted correction are not retained for this purpose.
What to expect
Recording takes roughly 20–35 minutes and can be paused. Possible risks include privacy loss, unauthorized access, discomfort from camera recording, and inaccurate model output. Safeguards reduce but cannot eliminate these risks. There is no guaranteed direct benefit.
Our commitments
We will not sell, lease, trade, or otherwise profit from your face or voice recordings, or data derived from them, as data, and we will never sell or license them to a data broker, advertiser, or other third party. This does not prevent us from charging for the product or service itself (for example, a subscription or API). We will not disclose your recordings to third parties except to the service providers who help us operate the research preview, with your consent, or when required by law. We will not use them to identify you or for facial recognition. We store them using a reasonable standard of care, and we destroy them when the consented purpose is satisfied or when you delete or withdraw. See the Privacy Policy for the full retention and destruction schedule.
Withdrawal
You can stop participating and request deletion. Withdrawal stops new collection and future use where technically and legally feasible but cannot undo aggregate analysis or model training completed before withdrawal.
The app and web record controls
The app or authenticated web recorder is the authoritative consent interface. It records the exact document version and hash, your individual choices, client build, locale, and acceptance time. If this public summary differs from the agreement shown before recording, pause and contact us.
Questions
Email privacy@dafdef.com or review the Privacy Policy.