Procedural fidelity is an important component of behavioral intervention programs. The "Train-to-Code" software was used to teach skilled observation of implementation of three types of discrete-trial programs, and improvement to procedural fidelity was assessed. Participants completed a training package that involved coding video examples and non-examples of a teacher delivering each discrete trial program. The degree of prompting given to the trainee increased or decreased dynamically during training sessions based on participants' coding accuracy. The efficacy of the training was tested within subjects via pre- and posttest role plays in which participants delivered discrete-trial programs to a scripted research assistant. Results indicated substantial improvement in discrete trial delivery at posttest. These results suggest that "Train-to-Code" may be an effective method for training delivery of discrete trial programs in applied settings.