To provide voice recognition and voice generation apparatuses which enable optionally setting of correspondence of a voice inputted to a voice to be generated for this voice, a toy with the same and a recording medium having voice recognition and generation control programs recorded therein.

A sound inputted from a microphone 10 is converted to a voice data by an A/D converter 13 and finally stored into a voice data memory section 15. A feature pattern of the voice inputted from a separated microphone 10 is extracted by a feature pattern extracting part 21, the extracted feature pattern is made to correspond to any of voice data stored in the voice data memory section 15 based on a direction from an operating section 3 and then, stored into the feature pattern memory section 25. As the result, when the feature pattern stored in the feature pattern memory section 25 is extracted again, a voice data corresponding to the feature pattern is read out of the voice data memory section 15 and finally emitted as sound from a speaker 20.