This paper presents our work on the development of a multimodal auditory interface which permits blind users to work more easily and efficiently with GUI browsers. A macro-analysis phase, which can be either passive or active, informs on the global layout of HTML documents. A subsequent active micro-analysis phase allows to explore particular elements of the document. The interface is based on: ( 1) a mapping of the graphical HTML document into a 3D virtual sound space environment, where non-speech auditory cues differentiate HTML elements; (2) the transcription into sound not only of text, but also of images; (3) the use of a touch-sensitive screen to facilitate user interaction. Moreover, in order to validate the sonification model of the images, we have created an audio "memory game", that can be used as a pedagogical tool to help blind pupils learn spatial exploration cues.