Techniques for automatically triggering video surveillance using embedded voice, speech, or sound recognition are provided. In one embodiment, a computer system can receive an audio signal captured from an area to be monitored via video surveillance. The computer system can further recognize, via an embedded recognition component, a voice, speech phrase, or environmental sound in the audio signal, and can determine that the recognized voice, speech phrase, or environmental sound corresponds to a predefined trigger condition. The computer system can then automatically transmit a signal to one or more video capture devices to begin video recording of the area.