At the time of this post, this was not possible. However, today specifying 'perFrame' as the outputStyle will give you results for each face for each frame.
https://dev.projectoxford.ai/docs/services/5639d931ca73072154c1ce89/operations/56f8d40e1984551ec0a0984e