With Unity wrote a similar Tom cat auto-detect speaking function

Simply add the script to the camera.

Here is the script I wrote, there is a problem because it is automatic call detection call recording Unity call delay will appear some delay small problem, you can refer to my code change to do a button click on the recording end play recording or can be achieved

Using system.collections.generic;using unityengine;using system.collections;public class tom2:monobehaviour{//publi    c float Power = 1;    public float min = 0; Public AudioClip audioclip = null;//audioclip Audio clip//public Audiosource audiosource;//audiosource audio source public Boo    L audioclipworking = false;    public string audioisplaying;    public bool microphoneworking = FALSE; Private list<float> sampleslist;    Play data store public float Samplesmaxvalue;    public int cliplength;    private void Start () {sampleslist = new list<float> ();///Not fixed startrecord ();        private void Update () {audioisplaying = Audio.isPlaying.ToString ();        Screen.sleeptimeout = Sleeptimeout.neversleep; if (microphoneworking) {if (!                Microphone.isrecording (null)) {microphoneworking = false;            Judgerecord (); }} if (audioclipworking) {if (audio.IsPlaying = = False) {sampleslist.clear ();                Startrecord ();            Audioclipworking = false;        }}} private void Startrecord () {audio.clip = Microphone.start (null, FALSE, 1, 44100);        while (microphone.getposition (null) <= 0);    Microphoneworking = true;        } public void Judgerecord () {float max = 0; Float[] Tempsamples;        Temporary data storage Tempsamples = new float[audio.clip.samples * Audio.clip.channels];        Audio.clip.GetData (tempsamples, 0);            foreach (float s in tempsamples) {float m = mathf.abs (s);        if (Max < m) {max = m;//The maximum value of the audio data just recorded}} samplesmaxvalue = Max;            if (Max > 0.1) {//Judge someone to speak Startrecord ();        foreach (float e in tempsamples) {Sampleslist.add (e);//Save Data}//Record another second }       else {//unmanned speech if (audio.clip! = NULL && sampleslist! = null) {  if (Sampleslist.count >) {audioclip myclip = audioclip.create ("Tom",                    Sampleslist.count, 1, 44100, false, false);                    Myclip.setdata (Sampleslist.toarray (), 0);                                       Audio.clip = Myclip;                    Audio.loop = false;                    Audio.pitch = 1.2f; Audio.                    Play ();                Audioclipworking = true;                } else {Startrecord ();            }} else {Startrecord (); }}} private void Ongui () {GUI.        Label (New Rect (1, screen.width, +), "Samplesmaxvalue:" + samplesmaxvalue); Gui.        Label (New Rect (2, Screen.width, +), "Cliplength:" + cliplength); Gui. Label (NewRect (3, Screen.width, +), "audioisplaying:" + audioisplaying); Gui.        Label (New Rect (4, Screen.width, +), "audioclipworking:" + audioclipworking); Gui.        Label (New Rect (5, Screen.width, +), "microphoneworking:" + microphoneworking); Gui.        Label (New Rect (6, Screen.width, +), "Audio.clip.samples:" + audio.clip.samples); Gui.        Label (New Rect (7, Screen.width, +), "Audio.clip.length:" + audio.clip.length); Gui.        Label (New Rect (8, Screen.width, +), "Audio.clip.channels:" + audio.clip.channels); Gui.    Label (New Rect (8, Screen.width, +), "microphone.getposition:" + microphone.getposition (null)); }}

With Unity wrote a similar Tom cat auto-detect speaking function