identifying, using one or more processors, a plurality of motion states and a control point of a computer animation model, each motion state of the plurality of motion states having a probability, the control point comprising a portion of a face or a portion of a body;

receiving audio data;

processing the audio data to determine a set of audio characteristics for the audio data;

determining a harmonic speed for the control point using the set of audio characteristics;

assigning a probability of zero to a subset of motion states of the plurality of motion states based on the set of audio characteristics;

pseudo-randomly selecting a first motion state that has a non-zero probability from the plurality of motion states;

generating a motion frequency value representing a frequency of repeating the first motion state using the set of audio characteristics and the harmonic speed; and

animating the computer animation model with the first motion state at the motion frequency value.