Enthusiast Transforms PNG File into Bird Call, Teaching Starling to Reproduce Sound Data

Enthusiast, bird lover, and musician Ben Jordan shared his successful experience of transforming a PNG file (176 KB of uncompressed data) into a sound signal and training a young starling to memorize and nearly accurately reproduce that audio information for subsequent conversion back into the original file. The experiment demonstrated that, theoretically, starlings can «transmit» data through their songs at speeds of up to 2 MB per second (in compressed form).

Jordan explained that he has been studying how birds respond to unnatural sounds and noise pollution for a long time. He has examined various mechanisms of sound production in songbirds and other species. For instance, the hairy woodpecker utilizes amplitude modulation, the northern cardinal has a broad frequency range, hummingbirds are difficult to record due to their skittishness, and the crested tit produces complex sounds with eighteen different tones.

Birds mimic the sounds of other species for various reasons. For example, the mockingbird mimics songs to attract mates and defend territory, the brown thrasher employs alarm calls to create a sense of security, and cuckoos mimic the calls of other birds to help their offspring thrive in strangers’ nests.

Jordan chose to focus on songbirds rather than parakeets because of their unique vocal apparatus – the syrinx – which allows for precise control over pitch and sound speed.

European starlings are known for being skilled imitators and intelligent birds. They can gather in flocks and produce noise, making them challenging targets for predators. To create sound, starlings utilize their counting skills. A solitary starling can mimic the flight pattern of the seven nearest birds. Their natural abilities enable starlings to distinguish and remember a wide variety of sounds.

Jordan revealed that he created a 176 KB PNG image and converted it into an audio file using a spectral synthesizer. He regularly played the resulting set of sounds to the starling, helping it learn the melody. Later, using an ultrasonic microphone and a Sony PCM-100, Jordan recorded the starling’s singing and decoded the results. It turned out that the starling was able to reproduce the image with remarkable accuracy.

«The bird reproduced the sound in the same range in which it heard it, effectively transmitting around 176 kilobytes of uncompressed data,» Jordan explained. When analyzing the audio signal from the starling, the original image can be discerned.

Interestingly, the sound signal was not natural for the starling’s range, yet the bird still successfully managed to reproduce this unconventional data.

Jordan estimated that with a compression ratio of 10:1, data could be transmitted and stored in the starling at speeds of up to 2 MB/s, albeit with different limitations.

Jordan noted that he was fortunate to have a starling named Roth for the experiment, who lived among humans and had been accustomed to unusual sounds from birth. The starling not only managed to reproduce these novel sounds using its vocal abilities but was also inherently adapted to them from a young age, making it ready for such vocal experiments.