Audio processing behaves differently across providers: audio may be transcribed, directly understood, or require a provider-specific format; if standard audio blocks are not supported, the format specified in the provider's documentation should be used, or the audio should first be converted to text and sent.