Animation Compiler

For addressing certain scientific questions, it may be beneficial to be able to present dynamic video stimuli continuously over a longer duration, with many variations of facial movement interleaved. The short animated videos provided in the MF3D Animations R1 stimulus set have been rendered in manner that allows users to concatenate multiple clips together to achieve this goal. Specifically, the first and last frames of each movie clip for a given head azimuth orientation are identical. Therefore, all that is needed to create longer movie clips containing many permutations of facial actions is to edit the original clips together in video editing software.

Video Editing in Blender

While there are many programs available for video editing, we recommend the video sequence editor (VSE) provided in Blender, for several reasons:

1. Blender is free and open-source under the GNU General Public License

2. Blender has an embedded Python interpreter, allowing Python scripting of various actions.

3. Blender is the software that we also used to render the images and movies of 3D macaque avatar, and hope to make publicly available in future.

Using MF3D_ConcatClips.py

MF3D R1 provides a limited set of movie clips, which are summarized in the accompanying spreadsheet (.csv file). MF3D_ConcatClips.py reads the data from this spreadsheet to find the appropriate order to position various movie clips in order to generate a new longer movie containing smooth continuous motion.

A screenshot of the Blender VSE window. Animated clips can be arranged manually or programmatically via Python scripting.

Example

For example, if a user were to specify the following variables in MF3D_ConcatClips.py:

ClipSequence    = ['Yawn','Rotate','Coo','Rotate','Scream']
HeadAzimuths    = [0, -60, -60, 30, 30]
PlaybackSpeed   = [1, 2, 1, 3, 1]

This tells the script to load the following sequence of five movie clips:

1. Load the movie clip in which the avatar performs a yawn action with the head azimuth angle at 0° (i.e. facing the virtual camera).

2. Once the yawn clip is completed, a subset of frames will be loaded from the head rotation clip, and appended in reverse order so as to produce a head rotation from 0 to -60° at a speed of 60°/second.

3. Load the movie clip containing a ‘coo’ vocalization with the head azimuth angle at -60°,

4. Load the frames from the head rotation clip to show the head rotating from -60° to +30°, at 105°/second

5. Load the movie clip containing a ‘scream’ vocalization with the head azimuth angle at +30°