Hi,
I would like to know if it’s possible (using the command line tools for the cross synthesis) to control dynamically the cents of the “frequency shape” parameter – the one that appears in the “generalized cross-synthesis” GUI – with the the use of a parameter file. Thus, if the answer is positive, would you be so kind to supply a complete command line example? Thanks for your kind concern.
Best,
Francesco Vitale
Dear Francesco,
I am a bit confused by this question. You refer to the generalised cross synthesis and whether the parameters can be controlled dynamically via command line. This is the case, as you may see even in AS we proposes to read a parameter file for dynamic cross synthesis. Then you ask about a frequency shape parameter, that appears in the constant cross synthesis dialog. Here the answer is yes but with a slight restriction. To understand this restriction you should first open the cross synthesis constant cross gui. The parameter that you can dynamically in a parameter file are amplitude scaling for sound 1 and 2, frequency scaling for sound 1 and 2 and the q factor. The parameter file you have to write consists of lines containing values for the time varying parameters X, x Y, y and q
time1 X_1 x_1 Y_1 y_1 q_1
time2 X_2 x_2 Y_2 y_2 q_2
…
As always in SuperVP , this file is read as a breakpoint function (bpf) for X, x, Y, y and q. These 5 bpfs for the five parameters are linearly interpolated over time.
I reproduce here the svp help setion on dynamic cross synthesis that explains ho these parameters are used
cross type cross: S1 and S2 are calculated in amplitude/freq representation and the output is
Amp(SO) = XAmp(S1) + xAmp(S2) +q*(Amp(S1)Amp(S2))^CE
Fre(SO) = YFre(S1) + y*Fre(S2),
the exponent CE is by default 0.5. It can be modified by means
of the flag -cross-exp.
YOu see that the amplitude of each bin of the output spectrum is generated by means of mixing the amplitudes of the two input spectra. Similarly the outout frequency of each bin of the output spectrum is a mix of the frequencies of the related bin of the two input spectra. Now these two mixes can be described in two different manners. You can either give the mixing parameters directly (what you would do in a parameter file), or you can describe the mixing process in terms of over all balance and gain. This would use the relations
X = A_B* A_G
x = (1- A_B) * A_G
Y = F_B * 2^(F_S/1200)
y = (1 - F_B) * 2^(F_S/1200)
The parameters that control this mixing process are A_B, A_G, F_B, and F_S and they are controlled by the sliders below the direct parameters and are named amp balance (A_B), ampl. gain (A_G), freq. balance (F_B), and frequency shaping (F_S).
So if you want the F_S parameter to change dynamically you can simply generate a cross synthesis parameter file as described in the beginning where you change y and Y using the eq here above.
Hope this all is clear.
Best
Axel
Ok Axel,
the question I posed may have been a bit confusing for you, but you understood exactly what I needed to do, i.e. to change dynamically the F_S parameter. Many thanks for your answer, which is, as usual, exceptionally clear and detailed.
Best,
Francesco
Dear Francesco,
In fact finally the question was not that confusing, but I started off after having had a look into the dynamic cross synthesis dialog where I did not find the frequency shaping. Then I changed to static cross synthesis and there I understood, but I forgot to remove the intro in me reply 
Best
Axel
Dear Axel,
unfortunately I’m having some trouble in reproducing the sound obtained by moving the F_S slider (with the aid of the sound preview option): in fact, when I use the parameter file (compiled on the base of your precious instructions) with the dynamic cross-synthesis, I get a totally different result. So surely I’m missing something. Moreover, the Forum won’t allow me to attach the files that document the differences between the “slider” and the “parameter” approach: can I send you these files by PM? Of course, I’ll wait for your answer here on the Forum. Thanks for your patience.
Best,
Francesco
Hi Francesco,
after looking into your parameter file I think you made an error in implementing the equations
X = A_B* A_G
x = (1- A_B) * A_G
Y = F_B * 2^(F_S/1200)
y = (1 – F_B) * 2^(F_S/1200)
A_B, A_G, ann F_B are in % (as shown in the dialog). F_S is in cents, this means you should divide the values by 100. This should bring you much closer.
You can easily see that you are wrong in putting example values into the AS dislog and see what are the corresponding values in the Amplitude Scaling and Frequency Scaling area (those are the values your equations have to reproduce) in the parameter file. If your parameter generating system works fine for a point in the parameter space it probably works everywhere. So you could try to reproduce a constant cross synthesis with a fixed parameter, your system should produce exactly the same result.
Note, that you also have to reproduce all the other values in the analysis and synthesis section as well as the Q exp (which should correspond to a -cross_exp flag).
The easiest to get all these parameters right is to run the const cross synthesis in AS and copy the flags from the command line besides those that are the arguments to -Gcross
Best
Axel
A_B, A_G, ann F_B are in % (as shown in the dialog). F_S is in cents, this means you should divide the values by 100. This should bring you much closer.
Sorry this reads confusing ;-). You should divide A_B, A_G, and F_B by 100. You most likely have got the F_S parameter right.
Best
Axel
Hi Axel, I divided A_B, A_G, and F_B by 100, and so I get a cross.par that looks like this:
0.0 1.0 0.0 1.6012147 0.0 200
0.25 1.0 0.0 1.4833511 0.0 200
0.5 1.0 0.0 1.4428032 0.0 200
0.75 1.0 0.0 1.4611682 0.0 200
1.0 1.0 0.0 1.5216574 0.0 200
…
The resulting sound (obtained with the command supervp -t -Z -U -norm -S"/Users/Francesco/Desktop/dynamic cross-synthesis/1.aif" -s"/Users/Francesco/Desktop/dynamic cross-synthesis/2.aif" -M0.5s -m0.166666999459267s -I441 -i1.47001483128406e-05 -Wblackman -whanning -Afft -afft -N32768 -Gcross “/Users/Francesco/Desktop/dynamic cross-synthesis/cross.par” -X0 -x1 -Y0 -y1.79929995536804 -q200 -cross_exp 0 “/Applications/AudioSculpt 3.2.7/Sounds/1-out 2.aif”) is closer to the sound that I get manipulating the F_S slider in the “constant cross” sound preview, but still different. Are there any other mistakes that I’ve made? Thanks for all.
Best,
Francesco
Hi Francesco,
First to ensure that you use the same (global) parameters, you can use the tip to hold the “Option” button in AudioSculpt during before you launch the preview mode of the cross synthesis. The command line will be printed to the console output so you can first copy-paste.
Then, – Axel will correct me if i am wrong – but looking at your command line i do not understand why you still have -X -x -Y -y params. I believe that if you use the parameter file you should not need to write again the X, x, Y, y in your commmand line, because the parameter file should contain already the “X x Y y q” parameters (according to the documentation of SuperVP for cross mode the file is made of lines with: time X x Y y q for add mode the file is made of lines with: time X x).
So the only extra parameter to properly set on the command line should be the -cross_exp.
HTH,
Best regards,
Charles
Sorry Charles,
I pasted the command line in the wrong way (inadvertently combining it with previously used command lines): here’s the command line that I actually use: supervp -t -Z -U -norm -S"/Users/Francesco/Desktop/dynamic cross-synthesis/1.aif" -s"/Users/Francesco/Desktop/dynamic cross-synthesis/2.aif" -M0.5s -m0.166666999459267s -I441 -i1.47001483128406e-05 -Wblackman -whanning -Afft -afft -N32768 -Gcross “/Users/Francesco/Desktop/dynamic cross-synthesis/cross.par” -cross_exp 0 “/Applications/AudioSculpt 3.2.7/Sounds/1-out 1.aif”. Of course the X x Y y q are absent, because they are specified in the cross.par
Best,
Francesco
Hi,
I managed to put the X x Y y q values in the right order of columns (on the base of the F_S and the A_S order of appearance in the dialog): now the resulting sound is much closer to the wanted result!
Best,
Francesco
Anyway, although the result I obtained is closer to the wanted sound (which can be produced by simply manipulating the F_S slider in the “constant cross” dialog), there are some strange things I cannot properly explain: in fact, while the sound preview in the “constant cross” dialog is normalized, the one in the “dynamic cross” is not (even if the “normalize output” option is checked). Why? Moreover, the difference between the wanted sound and the sound obtained with the parameter file (once I have executed the dynamic cross-synthesis command lines, including the normalization) is still very big. Maybe there’s still something wrong in the parameter file (here attached)? Thanks for your kind attention
Best,
Francesco
cross.txt (3.66 KB)
Hi Francesco,
there are 2 problem here:
If you use the command line then the problem is the norm flag. If you run the preview mode in AS then the norm flag is not respected (it can’t be because normalisation requires to see the complete audio file, which is incompatible with the preview mode). So if you remove the -norm flag from the dynamic command line, then you should get your result. The difference between normalised and non normalised sound hints at the fact that your cross synthesis output is clipping. This clipping is then part of the perceived sound, and obviously it changes the perceived sound.
If you use the AS preview mode with dynamic cross synthesis, then the norm flag handling will be fine, but it seems to me that AS does not set the -cross_exp flag in the dynamic cross preview this means you wont get the result there either). I am sure Charles will fix that quickly, and then you can use both command line and AS.
Best
Axel
Dear Axel and Charles,
I’m downloading now AS 3.2.8. Many thanks for the bugfixes!
Best,
Francesco