 This study aimed to test the ability of machine learning approaches to aid in detection of autism in self-recorded speech audio captured from children with ASD and neurotypical NT children in their home environments using three methods, random forests trained on extracted audio features, convolutional neural networks trained on spectrograms, and fine-tuned Wave 2 Vec 2.0, a state-of-the-art transformer-based speech recognition model. The results demonstrated that the models were able to predict autism status when trained on a varied selection of home audio clips with inconsistent recording qualities, which may be more representative of real-world conditions. The study suggests that machine learning methods offer promise in detecting autism automatically from speech without specialized equipment. This article was authored by Nathan A. Chee, Peter Washington, Aaron Klein, and others.