The system’s code and training data—drawn from one of the largest databases of speech recordings from autistic people—are openly available.