Confidence Scores

A confidence score is a value, often between 0 and 1 or expressed as a percentage, that an AI model outputs to indicate how certain it is about a particular prediction, such as a transcribed word, an identified speaker, or a detected tag. Confidence scores let teams flag low-certainty results for human review, filter or threshold outputs, and decide when a result is reliable enough to use automatically. They are common across transcription, tagging, detection, and separation workflows.