Dave Pawson's looked at self voicing SVG, I've attempted the same by extending my basic SVG image annotation tool, so that the generated RDF can be used for generating a description.
Currently, it's not technically self-voicing as the RDF,SVG and Image are all stored seperately, but they can be all embedded easily. The Text To Speech only works on Windows IE, and you'll need ASV, also if you do not have a TTS engine, then you will need http://activex.microsoft.com/activex/controls/agent2/tv_enua.exe and http://activex.microsoft.com/activex/controls/sapi/spchapi.exe which combined are under 2MB.
View and listen to the Talking description of Shepherds Image or a Talking description of the Tyndall photo (See Image Annotation ideas)
