Abstract: The disclosure provides a method for generating speech, an apparatus, a device and a storage medium, the method includes: generating an interaction intention of each node in an intelligent dialogue system according to an interaction description file, where the interaction description file includes node information of each node in the intelligent dialogue system, obtaining, according to the interaction intention and the interaction description file, at least one speech corresponding to each node in the intelligent dialogue system by using a generalization process, and storing at least one speech corresponding to each node, high efficiency of generating speech is achieved by automatically generating the interaction intention of each node according to the description file, thereby generating at least one speech of each node, and the speech is enriched through the generalization process, avoiding the problem that the speech is insufficient in the prior art.