VisionScribe Research

Privacy Policy

This Privacy Policy explains what VisionScribe Research ("VisionScribe Research," "we," or "us") records about you when you ask to take part and during your participation in this project, how we use it, who can see it, and how long we keep it. It should be read together with the Consent and Participation Agreement, which sets out what you are agreeing to.

1. Eligibility to Participate

You must be 18 or older to participate. VisionScribe Research does not currently accept participants who reside in Illinois, Texas, or Washington state, due to those states' specific biometric privacy laws. Full eligibility requirements, including the state of residence representation you make at enrollment, appear in the Consent and Participation Agreement.

2. If You Ask to Volunteer

Before you have an account, you can ask us for an invite. We store what you enter on that form: your first name, your last name if you give one, your email address, and how you heard about this project.

If we do not invite you within one week or you change your mind before an invite arrives, email contact@visionscriberesearch.com. If you do become a contributor, this information becomes part of your profile (see Section 8).

3. What We Record

While you read a sentence or phrase aloud, we record video of your face, including your mouth, jaw, and chin. We record your microphone at the same time, along with quality measurements, including how large your mouth appears in the frame, how bright the lighting is, and how loud the audio registers.

4. What Else We Ask You For

Your profile stores what you tell us about yourself: your first name, and optionally a last name, an email address, your age, your sex, your race, and the English accent you speak with. It also stores your state of residence, and whether you agree to receive promotional emails about this project. Only your first name and your state of residence are required. Everything else is optional, and you can change or delete any of it at any time from your profile page.

We ask for these details to measure whether the model reads one group of people as well as it reads another. A dataset that is uniform in age, sex, race or accent produces a model that works only for the people who happened to volunteer, and the only way to know that has happened is to have asked.

Your race, like your face, is information the law treats as sensitive — a special category of personal data under Article 9 of the GDPR. That is why we ask for it rather than assume it, why we collect it only with your explicit consent, and why leaving it blank costs you nothing.

Your state of residence is required because Section 1 depends on it. If you tell us you live in a state we cannot currently accept, we record only that fact — which state, and when — with nothing that identifies you, and no profile is created.

5. This Is Biometric Data

Your face and the way it moves can identify you. These recordings constitute biometric data and, therefore, require your explicit consent before we may collect them. You provide that consent through the Consent and Participation Agreement.

6. How We Use Your Recordings

We use your recordings to train a machine learning model to read lips from silent or whispered video, and to measure the model's performance. The audio serves only to confirm that you read the sentence or phrase shown on screen and to align the sounds with the video frames; the model itself never receives audio as an input.

We may license, sell, or otherwise transfer this dataset, in whole or in part, to third parties, including commercial partners such as video-conferencing or accessibility software companies. We may also use your recordings in future internal projects beyond the current lip-reading model. Any model or dataset derived from your recordings may be used, licensed, or sold commercially, by us or by parties to whom we transfer it.

7. Who Can See Your Recordings

Your recordings are stored in a private project operated by VisionScribe Research and its owner(s). Access is limited to you and the project owner(s), unless we transfer your recordings to a third party as described in Section 6 above. We never use your recordings to identify you or to search for you elsewhere.

8. Data Retention and Deletion

We retain your recordings while the project is active, and destroy them no later than five years after your last recording. If you request deletion, we destroy your recordings within 30 days of your written request, including any cropped copies derived from them.

This retention and deletion policy does not apply to copies already transferred to a third party before your deletion request. Contact us regarding data already shared with a partner.

Deleting your recordings removes them and anything derived from them that we still hold, but deletion cannot reach inside a model that has already been trained. Your data stops influencing the model at the next training run following your deletion request, not before.

Your profile information is kept while your account exists and is deleted with it. You can change or clear any optional field at any time, and deleting your account removes the whole profile.

You may delete your own clips, or request deletion, at any time from your profile page.

9. Security

We take reasonable measures designed to protect your recordings, but no method of electronic storage or transmission is completely secure. We cannot guarantee the absolute security of your recordings.

10. Changes to This Policy

We may update this Privacy Policy from time to time. We will notify you of material changes, and any updated terms will apply only to recordings made, or data practices occurring, after the effective date of the update, unless you separately agree to apply them retroactively.

11. Questions

If you have questions about this Privacy Policy or about data we hold about you, contact us through your profile page or through the contact information provided to you at enrollment.