Is it possible to compare two audio samples(meaning find word from one in another) realtime with cross-correlation?
No. If the words are pronounced with a different speed or intonation (prosody), or by different speakers, the cross-correlation won't have any significance.
A solution more robust to changes in prosody is to use audio features which are less dependent on pitch (such as MFCCs), and a comparison method that is robust to local time stretching (such as DTW or a distance between HMMs). It will still perform poorly when comparing a recording of two different persons saying the same word.
No. Human word recognition is a very complex psycho-perceptual phenomena. Things that sound like the same word to a human can have completely different waveforms with near zero correlation (different pitch frequencies and tonal modulations, timings, formant centers, surrounding context, etc.etc.)
Not possible... what @pichenettes says was right.. "People are not record players - every utterance of a word, even by the same person, has differences in timing, pitch, timbre which can alter the shape of the signal"
unless @user4947 is saying that it is the same audio file but maybe 1 is a cut out version while the other is full audio file..