Abstract: The present disclosure relates to adaptive auditory training systems and methods. A synthesized audio representation of an input text is generated and presented to a first user during a training session. Audio input from the first user corresponding to the input text is received, and a difficulty level of the synthesized audio representation is dynamically adjusted between consecutive prompts based on quantitative performance metrics derived from the received audio input. Acoustic thresholds are defined by signal-to-noise ratio measurements and response accuracy ranges, and difficulty adjustments may include modifying parameters such as timing, background noise level, word similarity, pitch, syllable count, or context clues. The synthesized audio representation may be personalized by training an artificial intelligence model on voice samples of a second user to create an emotion-agnostic voice clone preserving that user's speech characteristics.