Hi,
the title means to imitate a old practical IRCAM project:
An experimental SDIF-sampler in Max/MSP, http://recherche.ircam.fr/anasyn/rioux/publis/icmc02-draft-sampler-sdif.pdf
After 15 years cSound work in a more interactive way and it remains a solid open source DSP foundation. AudioSculpt can provide precise data than the model adopted.
The purpose is two-fold:
- explore the tonal dimension of resynthesis/cross-synthesis (with freedom)
- create an effective sampler implementation
.
In modeling a signal, it is of primary importance that the model be adapted to the signal in question. This leads to compact models which are useful for analysis, compression, denoising and true perceptual processing. The fundamental goal is for the recombination to be perceptually equivalent to the original signal with a transient and release management, and not for the synthesized residual to be a transparent version of the original residual.
Csound does not directly manage the SDIF format, apart Loris opcodes. Obviously you can do a conversion to a raw format for importing and write the algorithm from scratch but probably the ATS opcodes can be very effective for spectral time-domain resynthesis using a bank of oscillators.
ATSread facilitates individual control of the frequency and amplitude values of each one of the partials. As the unit works at k-rate, the frequency and amplitude data must be interpolated in order to avoid unwanted clicks in the resynthesis.
ATSreadnz is simpler than ATSread, because whilst the number of partials of an ATS file is variable, the noise data (if any) is stored always as 25 values per analysis frame each value corresponding to the energy of the noise in each one of the critical bands.
Perhaps there is someone in the group who may be interested in creating a link between AudioSculpt and cSound?
Best regards,
Elia