Add assertions for audio models

### Describe the feature or improvement you're requesting

Specifying a wrong model to audio transcription currently throws an error that doesnt make much sense. Asserting would prevent the users of the library making such mistakes.

### Additional context

_No response_