Embodiments of the present disclosure provide a method for improving speaker performance and a terminal device, when a speaker is being driven, an algorithm parameter that matches the speaker is selected, to resolve a problem that adjustment and control performance is reduced because of a difference in speaker models, and can effectively improve performance of a speaker. The method includes obtaining an identity of a speaker, where the identity is used to identify the speaker; obtaining an algorithm parameter corresponding to the speaker from a preset algorithm parameter library according to the identity of the speaker, where the preset algorithm parameter library comprises algorithm parameters of speakers of at least two models; sending the algorithm parameter corresponding to the speaker to a smart speaker power amplifier module, so that the smart speaker power amplifier module drives the speaker according to the algorithm parameter corresponding to the speaker.