Advanced Operation Options

By specifying the argument -behaviour_params=<VALUE>, you can change detailed conditions for the output animation curves.
Specify the path to a JSON file (UTF-8N) that follows the specifications below.

Available Parameters

The currently available parameters, their default values, and descriptions are as follows:

{
    "mouth_open_size_path" : "",
    "mouth_open_value_medium" : 0.6,
    "mouth_open_value_large" : 1.0,
    "mouth_open_value_small" : 0.4,
    "enable_adjust_delay" : true,
    "enable_targets" : [
        "PP",
        "FF",
        "TH",
        "DD",
        "kk",
        "CH",
        "SS",
        "nn",
        "RR"
    ]
}

Parameter	Description
mouth_open_size_path	Specify mouth size using a label file format. Specify the file path, e.g., 'mouth_open_size.lab'
mouth_open_value_medium	Maximum value for medium mouth opening. If the -enable_auto_detection_mouth_open option or mouth_open_size_path is not set, the animation curve is calculated within the range 0.0 ~ mouth_open_value_medium.
mouth_open_value_large	Maximum value for large mouth opening
mouth_open_value_small	Maximum value for small mouth opening
enable_adjust_delay	Whether to adjust the timing of the output animation to synchronize with the audio. If not adjusted, start playing the animation ahead of the audio by the number of seconds listed on the smoothing page.
enable_targets	Specify consonants to enable in the Mpeg4-Visemes animation curve file (.csv). Columns for consonants not specified will be set to 0.0, and vowel columns will represent mouth shapes. Vowels aa, E, ih, oh, ou are always enabled.

Example: Change Maximum Value to 1.0

To change the valid curve values to 0.0 ~ 1.0, specify a JSON file with the following content for mouth_open_value_medium:

{ "mouth_open_value_medium" : 1.0 }

Example: aiueo + n

For movements like "aiueo + n", specify a JSON file as follows.
PP and nn behave like the Japanese "n" sound.

If you use only the columns aa, E, ih, oh, ou and ignore PP and nn in the output animation curve, you can expect the mouth to close for the "n" sound in Japanese.
If your model has morph targets for PP and nn, you can use them as well.

{ "enable_targets":["PP", "nn"] }

Parameter: mouth_open_size_path

Specify mouth size using a label file format.
The first column is the start time (seconds), the second column is the end time (seconds), and the third column is the mouth opening level as a string.

0.0 1.0 large
1.0 5.0 medium
5.0 6.5 small

The mouth opening level can be specified with the following strings:

• large
• medium
• small

The tool option -enable_auto_detection_mouth_open provides a function to estimate this information from audio.