Hi,
I’ve created this 1GB7 SDIF in OM (starting from data obtained with an OS7 analysis), but, when I try to synthesize it I get a sound file with no duration: there must be an error in how the SDIF is written, but I cannot find it. Would you be so kind to tell me what am I doing wrong? Thanks for your concern.
Best,
Francesco Vitale
I see that the file hasn’t been attached. Here I try again.
Hello Francesco,
A few words of caution here - I am sorry for not having mentioned these notes when replying to your last question.
The GB7 format is for dumping the internal state of supervp, this format is not documented on purpose. It is meant to be written and read back by svp library only. The format is extremely complex, and there are numerous coherence constraints within this format - if you change one value you may need to change numerous others.
Modifying this format is a bit as if you would connect to supervp while it runs on your machine, change a few values in its internal memory, and then let it continue running. Any errors you do in the 1GB7 file can potentially put svp in an undefined state, which will then most likely lead to a crash.
For information: The intended use of this format is concatenation of preanalyzed sound segments - we don’t have means or plans to extend the use any further. Even if you use an existing 1GB7 sdif file and change only the content of the 1GB7 matrix internal data coherence may be broken, which may then lead to crashes. If you don’t mind this territory: try verifying your changes you did in OM with diff on txt representation of the original and modified sdif files, and see whether your changes are as intended. If they are, you made changes that are not supported by the internal state of supervp, if they are not, then you may fix your OM implementation, which then may lead you to a working sdif file (or not, in case the type of changes lead to internal inconsistencies)
With kind regards
Axel
Dear Axel,
thanks for your kind reply. You made clear what I already suspected: the values in this format are strongly interconnected, and changing one value determines somewhat of a dangerous chain reaction. Has there ever been a documentation for internal use at IRCAM that explains the coherence constraints of this format? Thant would be my only hope to understand fully how it works.
Best,
Francesco
Hello Francisco,
no such a documentation does not exist. This due to the fact that these interactions are not fixed. They depend on the settings that are used. Moreover, over time the relations may evolve and writing these down would first require a book of many pages and second hinder improvements of the internal algorithms, because they would have to fulfill additional constraints. I have to admit that I cannot tell you the relations of the head, I need to read the source code for this.
As I said already, the 1GB7 format is a dump of the internal state of the algorithm, and it is not meant to be worked on directly. To fully document it I would need publish the source code of supervp.
Best
Axel