You can get chips that you can record say 30 or 60 seconds of sound onto. You could use these with a prerecorded message for each bus. However every bus stop would have to be set up individually, very time consuming.
search rswww.com and you should fine them.
Alternatevely, you could construct a more universal system with a microcontroller, and hundreds of sounds stored in memory. When the bus comes the microcoltroller would say the number and destination. This is a more future proof design and a more flexable one. It would be more difficult to design in the first instance, but easier to roll out because of a universal system not needing customisation for every stop.
For a project, the ic recording chips will be sufficient to demonstrate the aim and system.