Around 2015 I was told about databending by a good friend, and we proceeded to give it a shot for ourselves to see what we could come up with. The idea of performing an audio manipulation such as adding an echo effect to an image sounded awesome to me, but there were issues we did not fully understand at the time.
We began by opening a JPEG (first problem) file in Audacity and chose some arbitrary values for the import process. This worked fine and allowed us to play back the JPEG data through my speakers, although it did not sound very nice at all. We performed some audio manipulation in Audacity (second problem, though this was the intended purpose of the experiment), and then saved the edited data as a MP3 (third problem) and renamed it ".jpg". The file would no longer open!
Data compression algorithms such as JPG (for compressing image data) or MP3 (for compressing audio data) are wildly different from one another, and not all data are valid compressed files. In order to view a JPG file or play a MP3 special software is required. Fortunately most modern Operating Systems such as Microsoft Windows, Apple macOS, Android, GNU/Linux, FreeBSD, and others include the necessary tools.
If you were to edit a few random bytes inside a JPG file, it is possible that you will corrupt the contents in such a way that the file will no longer be valid and viewable. This is because the decompression algorithm used when you open the JPG to look at it will "expect" a certain structure to the data, and not all values are valid at all locations.
When we performed an audio manipulation on the JPG data, we corrupted the structure of the data in such a way that the result was no longer valid as a JPG, and could no longer be viewed.
We laerned something from our failure, but didn't touch the project again for another four years, until Experiment One: Edit Image As Audio.