Create a sound from a image?

#1

Did anybody thought about a solution to create a sound from image data? I think the idea is quite interesting. A image has less data than a sample. I saw some imagesynths in the web and tried some. I think the most solutions were not very successful, but the idea behind is not bad, depending on the implementation.

#2

I'm glad someone else is interested in this!

I'm not sure the Axoloti is powerful enough to do this usefully.

I think a Raspberry Pi would do the job better, and you could also use the RPI camera module to snap images to resynthesise.

a|x

#3

The wave/play objects just read raw data, so if you would put some images on the SD card you can just play them as if they were audio. You should expect a lot of noise though when playing files like this. Works with any kind of binary file, so Excel and PowerPoint files do a fine job as well.

I've mainly been using Photosounder to convert images to sound: http://photosounder.com/

#4

Ah, I was thinking more of spectral resynthesis, where each line of the image represents the amplitude envelope of a sine wave of a particular frequency.

That's what Photosounder does, I think, but it's a bit different from just playing an image file as audio.

a|x

#5

Photosounder indeed interprets each line of pixels as sound information, and I think this might be hard to do on Axoloti.

#6

It might be possible to create a sine-bank patch like the Nord Modular one, but i imagine it would be quite low-res (ie not that many sines).

a|x

#7

Yep, this has been thought a lot by many. Have you heard of the ANS synthesizer?


There is a very nice implementation here:
http://www.warmplace.ru/soft/ans/
I do not think you can simulate this with the Axoloti, the ANS has a bank with lots of sine waves.

#8

Yeah, i requested a sine bank object some time ago, like reaktor's one which has 256 particles iirc. Ability to build something like Razor or Lazerbass with Axoloti could be awesome. This also could be used for resynthesis a'la ANS and pvoc-alike sound generation.

#9

I think that might be a bit ambitious, since these Reaktor synths are quite processor-intensive, even on a desktop machine.

The other issue with additive-type synthesis is actually coming up with some useable way to control all those sine waves. While it's relatively easy, technically, to generate a load of waves, it's much more difficult, UI-wise, to come up with a way to sculpt these into something interesting.

Razor's fancy 3D graphics help a lot with that, but obviously, that's not an option with Axoloti.

a|x

#10

I guess we can look at some hw additive implementations like Kawai K5000 or Kurzweil K150

#11

There's also the Yamaha FS1R, which (I think) did additive synthesis.
They never really took off though, maybe because they're hard to program.

a|x