As described above, various audio characteristics can be determined from audio data. In one example, an audio energy value or a “danceability” value is determined for a segment of audio data, and this can be used as a threshold to determine whether to generate a pseudorandom animation matched to the audio data. Audio data 680 of FIG. 6C, for example, can be determined to have a tempo similar to a tempo of audio data 670 of FIG. 6B. However, the additional details (e.g. frequency content, beat strength, beat consistency, etc.) can be used to determine additional characteristics, that can be specific to a framework or particular computer animation model and definable by a designer. In the example of FIG. 6C, audio data 680 does not meet a threshold for animation during time periods associated with no motion states 671 and 675, but meets the threshold criteria for motion during the time period for motion states 673 and 679. As described above, motion state 673 and motion state 679 can be selected randomly from the state-space for the computer animation model being executed based on the assigned probability for all motion states, including selected motion state 673 and selected motion state 679 during the time periods when those motion states are used to generate a video animation. In some embodiments, during no motion state 671, the computer animation model is frozen, and does not move at all in the video frame. In other embodiments, a default or “waiting” animation can be used, that transitions to a selected motion state when an audio data threshold is met. Such embodiments may use a transition animation between such states, or can move directly between the states. Similarly, in various embodiments described herein, the randomly selected motion states can have transition animations used as the computer model transitions between the motion states, or can simply animate movements based on a current motion state as new motion states are selected.
