For the input file I selected a short transient (a snare i'd made from my own library)
For the Impulse Response I selected a long granular choir sample from a free pack downloaded a long time ago.
The results were not impressive. I made the assumption that a long pad sound would make a good impulse for the short transient input file.
I reversed the sample selection for the input file and impulse file to test whether this would have some effect on the output.
I was disappointed to discover that it didn't.
Clearly I have a lack of understanding of the convolution process. After taking a look at the FScape Module documentation I took another approach with sample selection and experimented with the settings.
| "the spectral characteristics of the IR (both amplitude spectrum and phase resonse) are superimposed onto the input file. When using normal sounds the output tends to emphasize the low frequencies unless the IR has a flat spectrum. It tends to emphasize the frequencies that exist both in the input and the IR and cancels those frequencies which are not present in either file." |
I opted to use a dynamic sample from my library as the impulse response with a sample of granular drone textures from a sonified image as the input file. I also experimented with the convolution with inversion mode, morph IRs with correlate and shift policy setting along with Truncate overlaps which the documentation stated caused "interesting" results from buggy code.
Input file: granular drone textures from sonified image
Impulse file: dynamic digital transient + fx sample from personal library
Output file: Convolution with inversion mode, morph IRs with correlate and shift policy setting along with Truncate overlaps
Output file: Convolution with inversion mode, morph IRs with Polar Crossfade policy setting along with Truncate overlaps
Output file: Deconvolution mode, morph IRs with Polar Crossfade policy setting along with Truncate overlaps
Output file: Convolution mode, morph IRs with Polar Crossfade policy setting along with Truncate overlaps
Interesting method for generating textures when superimposing rhythmic content. I would like to use recordings of speech patterns as a source for future experiments.
Finally I decided to use an impulse response from a range I recorded myself last year at First Site Gallery at RMIT for a previous assignment. For these experiments I used dry sample of a Koto playing Kojo no Tsuki as the input file and a mouth click impulse rather than one of my balloon pop impulses.
Input file: Dry sample of Koto playing Kojo no Tsuki (The Moon Over the Ruined Castle)
Impulse file: Mouth Click + Impulse Response recorded at RMIT First Site Gallery
Output file: Convolution mode
Output file: Convolution with inversion mode + truncate overlaps
Output file: Deconvolution mode + truncate overlaps
Great results, especially with deconvolution mode. I have used convolution reverbs in this way in the past, but not with this utility. FScape has some interesting quirks I enjoyed exploring. I will definitely be exploring the rest of its modules and experiment with convolution further with other rhythmic and textural combinations.




No comments:
Post a Comment