Commit 70006912 authored by Greg's avatar Greg


parent e2fd629f
AudioGuide is a program for concatenative synthesis that I'm currently developing with Norbert Schnell, Philippe Esling and Diemo Schwarz. Work began in 2010 at IRCAM when I was composer in residence for musical research. Written in python, it analyzes databases of sound segments and arranges them to follow a target sound according to audio descriptors. The program outputs soundfile event lists that can be either synthesized (in csound or Max MSP/Pure Data) or translated into symbolic musical notation.
AudioGuide is a program for concatenative synthesis developed by Ben Hackbarth, Norbert Schnell, Philippe Esling, and Diemo Schwarz. It is written in python, however, do not need to know python to use AudioGuide - the user supplies simple options files that are written in python's syntax to interact with the program.
AudioGuide differs from other programs for concatenative synthesis in three important ways.
AudioGuide should run out-of-the box on any recent OS X with python2 or python3. The python module numpy is required, but comes preinstalled on most newer versions of python. If you want to have AudioGuide automatically render concatenations, you need csound6 installed as well.
1. Database samples are matched to the target sound based upon time-varying descriptors, thus permitting similarity to account for sounds' morphology. This allows for longer chucks of the database to be matched, enabling concatenated outputs which used whole acoustic sound segments rather that windowed grains.
2. The user may include, exclude and/or weight various audio descriptors to achieve different similarity contexts (rejecting the idea that a context-independent notion of similarity exists), such that a single target and a single database may be used to create a vast array of "variations" in likeness and semblance.
3. Samples can be selected simultaneously (either vertically or hoizontally overlapping) to contribute to the representation of the target's characteristics. Therefore, several corpus samples might can be used to approximate different aspects of the target's energy. In addition, database samples can layered to better approximate the entirety of a target sound, if so desired. To hear these results, look at the simultaneous selection examples on the examples page.
AudioGuide differs from other programs for concatenative synthesis in several notable ways:
Here is a complete list of the resources that AudioGuide1.35 requires on your computer:
* csound6 - Needed if you would like AudioGuide to automatically render concatenations (which you probably do). Download an installer from here.
* numpy (python >= 2.7 has numpy pre-installed) - Numpy is a numerical computation plugin for Python. Upgrading to the latest python2.7 will automatically install numpy. If you don’t have it, you can download the source code or a binary installer here.
* matlibplot (very optional) - Install this python module to enable graphing of descriptors and soundfile segmentation. Get it here.
\ No newline at end of file
* AudioGuide is not realtime and therefore sounds can be layered much more densely compared to realtime concatenation. Non-realtime analysis also permits more flexible and creative mapping between target and corpus descriptors as well as algorithmic accounting for overlapping corpus sounds in descriptor calculations. More info about how to control the superimposition of sounds is here.
* AudioGuide gives a large number of controls for fine tuning what sounds are included in the corpus, permitting the user to include and exclude segments according to descriptor values, filenames, restricting segment repetition, scaling amplitude, etc. See all of the options here.
* AudioGuide aims to give maximum creative control over how the sounds of the corpus are mapped onto the target. Many different configurations for normalizing corpus and target data give the user a higher degree of control over the results and permit creative flexibility in defining similarity.
* Similarity between target and corpus sounds can be evaluated using time-varying descriptors, thus giving a better sense of the temporal morphology of sounds. A full list of available descriptors is here.
* AudioGuide has a robust and flexible system for defining how corpus samples are matched to target segments. One may find the best match according to a single descriptor, but one may also define multiple search "passes", effectively creating a hierarchical search routine. One may also create boolean tests within the search function to further nuance the search process. See here.
* By default AudioGuide creates a csound score that is automatically rendered at the end of the concatenative process. However the program is also capable of creating a json output, which can be played in Max (patch provided in the distro), textfile output as well as midi output.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment