Speech synthesis method and speech synthesis system

The invention provides a speech synthesis method, which comprises the following steps of: 1, preparing an audio material library, dividing audios in the audio material library into multiple classes according to types, and labeling the audios; step 2, training the labels of the audio materials obtained in the step 1 based on a Bert depth model, and obtaining word vectors corresponding to the audio materials; 3, inputting a text fragment of the novel, and performing paragraph structured analysis on the text fragment to generate a structured drawing book; 4, according to semantic approximation, obtaining different types of words in the structured picture book and audio candidates with the highest similarity in the audio material library; and 5, calling an applicable TTS engine, fusing the matched candidate audios in the audio material library, and performing voice output according to the required output structure information. As a novel technology, the method can integrate the TTS engine and the gender and emoti