I have to admit that I intentionally didn't do a search before I posted this because it seems so far fetched (in my mind). That being said - here goes. So, I know very little about compression algorithms, DSP, programming and everything else that would be required to know to make this work even conceptually, but here is the idea:
Many audio player software packages come with display plugins to make music as pleasing to watch as it is to listen to. I believe those visualizations are programmed to read from the audio source and make a pattern or color on the screen that changes over time based on how the sound changes.
What if you were to play your audio through your audio player software of choice and record a video (no audio) of that visualization. Then, compress the video with a high efficiency video codec (a la H.264, xvid, etc). That compressed video would be your audio data, and you could easily control your audio resolution by the pixel resolution of your video file. You could then reverse the process with the video file as the input to the decompression algorithm that takes colors and patterns from a video file and converts them into an audio signal.
I know this sounds like a really big stretch, but wouldn't it be cool to pull off? It would be interesting to see how the visual compression artifacts would convert to audio artifacts.
Any thoughts? Has this been done before?
I can't find the link right now, but this has been done before. In particular, audio data was encoding using JPEG, and image data was encoded using MP3. The results were much better than expected (since the expectation was pure noise 
, but not nearly good enough to warrant additional research.
, but not nearly good enough to warrant additional research.
Lets see... Take a 1.4 mbit audio stream, ignore most of the information in it to generate, say, 8 mbits of only slightly related video information, then throw away most of the detail in the video stream. Now expand out the video data and try to figure out what audio data might produce that pattern. Sounds like a plan to me!
QUOTE(Hunkadoodledoo @ Nov 20 2007, 07:24) 
What if you were to play your audio through your audio player software of choice and record a video (no audio) of that visualization.
...
You could then reverse the process with the video file as the input to the decompression algorithm that takes colors and patterns from a video file and converts them into an audio signal.
...
You could then reverse the process with the video file as the input to the decompression algorithm that takes colors and patterns from a video file and converts them into an audio signal.
It is not possible to do the reverse transformation. Most of the audio data is lost.
Leave alone H.264 and other video coders. Can you at least reconstruct the audio from the uncompressed vis capture? I don't think so.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.