Welcome Linda,
The only way to do this in the API is to write them out as a WAV and submit them to runRecognitionOnWavFileAtPath:(NSString *)wavPath usingLanguageModelAtPath:(NSString *)languageModelPath dictionaryAtPath:(NSString *)dictionaryPath languageModelIsJSGF:(BOOL)languageModelIsJSGF;
However, it should be pretty quick since it doesn’t need to calibrate.