Comment This is actually old tech. (Score 3, Insightful) 62
The ability to convert a spectrogram to sound has long been known in the speech research community. In 1950 a device known as the Pattern Playback was built at Haskins Laboratories. You would draw an artificial spectrogram and feed it to the machine and it would play back the corresponding sound. It was used to perform experiments on the acoustic cues for speech perception. The original machine was last used for research in 1976.
See the Wikipedia article: https://en.wikipedia.org/wiki/...