I03.resynthesis.pd patch is pretty much everything I need for my current sound design project, but the problem is that I need to automate it... So how do I...?

1. Store/recall different curves from the table? Can I save these as text files or something?
2. Morph between these stored curves at a certain speed?

The idea is that I use this principle to shape the spectrum of white noise, and then use bangs or [metro] to switch between these stored spectrum curves. It should smoothly morph between curves at a pre-set speed (using [line] perhaps?). In a similar fashion to GRM Tools preset morphing. That would be superb!

Any advice / implementation examples would be very much appreciated!

THANKS IN ADVANCE!

1. help->2.control.examples->15.array for how to save the curve to a textfile.

2. [tabreceive~ table1] [tabreceive~ table2]
   | |
   | [r crossfade] | [r crosfade]
   | | | |
   [*~ ] | [- 1]
   | |
   | [abs]
   | |
   [*~ ]

then join those [*~ ] objects together where the original [tabreceive~ ] is in the resynthesis patch

Thanks for your incredibly quick response, hardoff!

1. curve->textfile->curve works perfectly! Incredible!

2. But I'm not quite sure where you mean the original [tabreceive~ ] is in the resynthesis patch... [tabreceive~ $0-gain] in the fft-analysis window I assume? But where exactly do I connect those [*~ ] objects? I tried several locations, but it doesn't seem to do anything... Also how do I trigger them? Where do I send my bangs? Also, what defines the speed of crossfade?

I very much appreciate your help, sorry if I sound dumb or annoying with these questions.

THANKS THANKS THANKS!

@vytis said:

2. But I'm not quite sure where you mean the original [tabreceive~ ] is in the resynthesis patch... [tabreceive~ $0-gain] in the fft-analysis window I assume? But where exactly do I connect those [*~ ] objects?

Try the attached, has crossfade slider, but you could equally use [line] to control the morphing.

General technique would be to:

1. fade from table 1 to table 2
2. load something new into table 1
3. fade from table 2 to table 1
4. load something new into table 2

and repeat

http://www.pdpatchrepo.info/hurleur/I03.resynthesis-crossfade.pd

This is amazing, thank you!

I don't want to sound like I'm nit-picking - your patch is close to what I'm trying to achieve, but is it possible to actually *morph* the frequency curve, rather than crossfade the volume...? So that each frequency band adjusts to a value from table2?

MANY THANKS!

Claudius, this is indeed a very interesting patch with a very nice explanation inside.

@ Vytis : i might be wrong but isn't crossfading the same as "morphing" since each band (point) in each table is "amplified" more or less depending on the crossfade-slider, which goes from 0 to 1? Or how would you define morphing?

domien - I guess that for white noise the result of "morphing" or "crossfading" will probably sound very similar. So perhaps you are right.

... But what I'm thinking about is that the process of morphing should also affect frequencies inbetween. I've drawn this quick sketch to illustrate it:

Red curve is table1, Blue is table2, Black - transition. As you can see, it affects frequencies in between which I don't think is the case with crossfading.

Such morphing possibility would make PD a most powerful sound design platform for synthesising very complex sounds in a rather user-friendly way. I came to this conclusion when altering curves in real time, it produced some very realistic bird call sounds, car brake squeaks and creaks etc. I don't think that this can be achieved with regular crossfading.

Am I asking too much?

I'd like to take this opportunity to say that this forum is simply amazing! I've learnt so much in the last couple of weeks. It has encouraged me to actually start using PD for my sound design work. I'll certainly share my patches as soon as I develop something interesting.

Thanks guys!

You certainly are asking for a smart algorithm. describing your two curves algebraically, the red curve is a x^6 curve, whereas the blue curve is a mere x^4 curve. The two highest humps (in terms of x val) transition to the two humps of the blue curve. But what happens to the low hump of the red curve? It simply recedes.

For this kind of decision making, I'd probably dy/dx the two graphs and compare the number of sign changes, and then manipulate the graph derivatives on a third and fourth tables to develop the swell respecting morph.

BTW- the graphic example you drew is real swell. May I ask how you generated it? perhaps if it was automated, you could write each successively to array, or pass the info over a socket?

http://www.pdpatchrepo.info/hurleur/fftfade.pd

sunji, this patch is precisely what I was looking for!!! Thank you so much! Wow!

I used vector graphics package Xara to create the illustration. Its blend tool automatically generates shapes between objects. It's open-source and therefore completely free for Linux, however you'd have to pay for its more advanced version for Windows. If you are interested:
http://www.xaraxtreme.org/
http://www.xara.com/

This forum is amazing!
Thank all of you guys so much! Thanks hardoff, ClaudiusMaximus, domien and sunji! Your help is very much appreciated.

you make this forum amazing, vytis! You arrive to the forum with questions and soon enough you find yourself answering them for the new arrivals. Isn't this world a great place for knowledge seekers?