Hi !

I am trying to do something, and have no clue how to do it (google didn't help) :
My goal is ... on a website that I am writing, when people click some stuff, I want the web-server to quickly generate a sound, and send it back to the page so that it can be played to the user. So the generation of the sound has to be as quick as possible.

an example - I have a very simple patch :

[receive x1] -> [osc~] -> [dac~]

And I would like to programatically (i.e., not using the GUI) generate say 5 seconds of sound of this patch. And where it becomes more complicated, is that I would like to change the osc's frequency at given time (for example, at 0 send 440, at 1.2 seconds, send 880, ...).

Of course, in order to send the file as quick as possible, the generation shouldn't be real-time like it is when using the GUI. So basically, I want to use pd as a calculating machine (unaware of time), writing directly to a .wav (or whatever file format).

Does someone know if this is possible ? Or has any suggestion ?

Thanks and cheers !

I may be doing something similar soon. Here's what I was thinking:

Have the web interface write data to a .txt file and then run Pd from the command line (using a shell script, I guess). The Pd patch that you're using will open the .txt file and record the .wav file, then close.

I think every part of that is doable. The website/server side of it is more the unknown territory to me...

Hi acreil !

Pb is that if I do that, it will record the .wav in realtime ... so if I want 5 secs of sound generated, I'll have to wait at least 5 seconds, which is not an option in my case !
Or do I miss something ?

As for writing the data to a text file (I guess you were talking about the frequency changes), then open the file from pd, it does sound like a good solution ! Though I am not sure how to achieve it in practice : I am not very familiar with reading/parsing files in pure data , and I am not sure how I can encode the information time/message in a file so that is would be easy to decode with pd. Any idea with that ?

Hey, what's about using data-objects (no "~") to write into an array?!
Or maybe you can set a higher samplerate (...what's pds max samplerate??) in order to get things done faster. ...won't work with objects that calculate time in relation to the samplerate..

You'd be running it without the GUI or audio, so it would run as quickly as possible. You could use [textfile] to import the data, but the formatting is up to you.

just a question, but do you really NEED to have the file saved as a .wav file??? If you're just going to stream the sound through your webpage, maybe there is a way to take pd's [dac~] output and directly route that to some other streaming audio format???

of course, i have never tried this myself, and there may be serious issues which make this difficult, but in theory i can't see why it wouldn't work.

@hardoff said:

maybe there is a way to take pd's [dac~] output and directly route that to some other streaming audio format???

That I have tried before I have my own virtual server hosted by some company, and I had set-up :

• an icecast server : which basically takes some audio and streams it
• darkice : a client for icecast -> icecast server needs a client to feed him some audio, and darkice is a good one since it can be connected with jack !
• a super-collider server : running just a simple oscillator and connected with jack to darkice.
• a webserver : serving a page on which I can listen to the streamed audio, and send an ajax request to change the frequency.

The audio was basically doing that :

jack icecast protocol
SUPERCOLLIDER -------------> DARKICE --------------------------> ICECAST ----------> INTERNET

And it worked well ! I could basically control the supercollider server from the internet ! Thing is, there was a veeeery big latency : at least 10 seconds between the moment the frequency changes, and the moment I hear the change from the audio stream. This is inherent to internet streaming and inherent to streaming in general, since streaming techniques focuses on not cutting the stream rather than giving it instantly to you (and therefore needs a shitload of buffers everywhere). I could have tweeked it a bit, and maybe have a smaller latency, but it would have been at least a few seconds - so no good.

My idea this time is rather to fake a direct control with webpd (which is really memory-greedy on your browser), meanwhile the server generates the sound, send it back, and when it is ready to be read, it starts and webpd stops (to save memory).

I have actually succeeded now. You have to use the option "-batch" and it generates the sound instantly which is what I was looking for ! For some reason, I also need to disable realtime-priority, but that might be because my ubuntu is not well configured. So I use :

pd -nrt -batch mypatch.pd

Awesome! Thanks for making me try this batch stuff...
I didn't think it would work cause I found this:

http://eat.adbk.de/index.php/Pure_Data_Configuration_Flags

...and since I use windows..

But, it works like a charm! (don't forget to put a "[ ;pd quit(" at the end )

You're welcome !

> http://eat.adbk.de/index.php/Pure_Data_ … tion_Flags

Why is it stroked ? Does somebody know, is it deprecated or something ?

> don't forget to put a [ ;pd quit]

Hey Thanks a lot for the tip ! I was actually googling to find how to exit pd, and I didn't find it ! So thanks a lot