Table 1. Normalized Root-Mean-Square Error (RMSE) of TorchDIVA motor signal.

Production Label	Training	Trained	Maximum
happy	0.08%	0.11%	0.12%
i	0.06%	0.08%	0.09%
u	0.06%	0.08%	0.09%
e	0.06%	0.08%	0.09%
ae	0.06%	0.08%	0.09%
example	0.06%	N/Aa	0.09%

Average normalized root-mean-square error (RMSE) in motor command output between TorchDIVA and DIVA over 20 repetitions of the listed speech production. RMSE was measured under both training and well-trained speech target conditions. The final column contains the maximum observed RMSE over the 20 repetitions. The absolute difference between the two signals is expressed as a percentage of the magnitude of the signal range.

^aThe production labeled “example” has no default pretrained forward motor program.