Targeting the problem of inconsistent volume between host and guest during podcast recording, the Leveler app uses reference audio to drive the model to generate an audio engine, compressing the development cycle to the half-hour level.