Here is provided basic code to accompany the following paper:

   Cans and Cants: Computational Potentials for Multimodality with a Case Study in Head Position
   Rob Voigt, Penelope Eckert, Dan Jurafsky, and Robert J. Podesva.
   2016. Journal of Sociolinguistics.

As written, it expects a 'videos' subdirectory with video clips you wish to analyze. (Note that it expects videos like those described in the paper, in which the subject's head is in reasonably full view and reasonably pointed towards the camera.) It produces a dataframe (data.csv) with phrase-level measurements of a number of acoustic features as well as estimated head cant angle (mean and standard deivation).

I'm glad to help with any questions if I can, please just e-mail me. This code will be a small piece of a larger and more cleanly prepared software package for automatically generating dataframes from audio/video data with a number of visual variables and at a number of scales of analysis, so keep an eye out for that!