There is disclosed a system and method for addressing an audio file server to play pre-recorded audio files, including variable audio files, using a query URL containing the required file's attributes, without requiring a fully-resolved file address. The HTTP URL protocol is used by adding attributes, such as the language, the speaker, and a text version of the desired message, along with other required attributes of the audio file to the URL. The audio file server accepts and analyzes the attributes in the URL to find out what type of variable field is being requested. Normally, variable field prompts created from spliced audio clips are restricted to a few specific types of variable fields, such as time, date, or amount, fields, or numeric strings such as telephone numbers, credit card numbers, etc. Once the audio file server determines the field type, language and speaker from the URL, it examines the field text value from the query attribute string. The file server then calculates and retrieves the set of utterances required to create the desired phrase. The audio file server splices all of the short files together, and returns the completed utterance to the voice browser for playing to the user.