An audio/image converter using a raw approach. A byte-to-byte translation is provided between image file and audio files (in both directions), in order to misuse image and sound processing algorithms.
Using this tool before and after using a Digital Audio Workstation (such as audacity, ardour or reaper) gives a way to apply audio filters on images.
Using this tool before and after using a Raster Graphic Editor (such as gimp) gives a way and to apply image filters on audio files.
This tool is inspired by Cécile Georges' artistic practices, and aims to extend the possibilities offered by digital bending.
The raw conversion uses a byte to byte conversion, each audio sample being translated into a pixel. By consequence, we obtain a time wrapping, each audio sample being the neighbor of a sample from the past (and future). The size of the wrapping (in ms) depends on the audio bitrate and the width of the image.
Note: stereo audio files and RGB/RGBA images are not described in this image, but it works in a similar way (each audio sample is a 2-bytes data, one per channel, and each pixel is a x-bytes data, one per color).
sudo apt install python3-pydub python3-pil python3-appdirs python3-pyqt5
- Add dependancies (see below)
- Download this repository
- run
./install.sh
Without spectific option, the script is using history to adjust the properties of the output. For example, in the following sequence, the first run produces a stereo 44.1 kHz audio file (default format), and store in the history the specific properties of the input image (width, height, RGB/RGBA). The second run uses the history to identify the correct image parameters, using the properties of the audio file (number of samples, number of channels and sample rate) as a filtering to identify the possible configuration of the audio file ancestor.
rawdodendron.py -i image.png -o audio.wavto convert an image to an audio filerawdodendron.py -i audio.wav -o image.pngto convert an audio file to an image
You can of course force properties using command line parameters:
rawdodendron.py -i audio.wav -o image.png -w 300 --rgb
All the command line parameters are visibles using the following command:
rawdodendron.py -h
A Qt5 interface is provided (only available in french), in order to process a series of conversion, editing each of the available parameters for each input file.
Right clic on an image file and find the rawdodendron entry to convert it to an audio file.
Right clic on an audio file and find the rawdodendron entry to convert it to an image file.
Here is a link to a Youtube video illustrating the use of image processing algorithms on an audio file. Description is given in french.
Some images generated from an RGBA png version of the initial rhododendron image, using the following process:
- convert the image to an audio file
- apply an audio processing algorithm (compression, reverb)
- convert back to an image
The last image is the result of a reverb, but from an RGB image (without alpha channel):
