Inference & Profiling · Company practice
Videos, images and audio are analyzed for objects, faces, spoken words and demographic classification
TikTok collects content characteristics and features of the videos, images and audio recordings in your user content, for example identifying objects and scenery, the existence and location of a face and other body parts, and the text of words spoken in it. It says it may collect this to enable special video effects, for content moderation, for demographic classification, for content and ad recommendations, and for other non-personally-identifying operations.
Details
- Who
- TikTok
- What it covers
- objects, scenery, faces and body parts, spoken words in user content
- Why
- video effects, content moderation, demographic classification, content and ad recommendations
Supporting quotations
Content characteristics and features, including about the videos, images and audio recordings that are part of your user content, for example, by identifying objects and scenery; the existence and location of a face and other body parts; and the text of words spoken in your user content.
We may collect this information to enable special video effects, for content moderation, for demographic classification, for content and ad recommendations, and for other non-personally-identifying operations.
Qualifications & exceptions
- Purposes are introduced by "may", and the list ends with an open-ended "other non-personally-identifying operations".
Cross-references
- qualified by (model-proposed): Faceprints and voiceprints may be collected from your content; permission is sought only where required by law — The biometric card adds that faceprints and voiceprints may be collected from user content, with permission sought only where required by law; this limits the general content-analysis card.