So I have a week to get this sorted (I hope).

I'm writing my live set in PD and have 9 tracks with 5 stems each. I've tabread4 set up so that I can play, pause, rewind and fastforward my tracks to cue them and such.
I've come across an issue however. If I load all the 45 stems in then it overloads the RAM and PD crashes. If I try and load the audio in while another one is playing, the audio will cut out while the new ones are loading.
If I add all the stems before playing with loadbangs and delays then it will crash when it's all loading.

Is there anything I can do at all about this (other than get a more powerful laptop)? I really don't want to go back to Cubase

Thank you

there's a workaround to the dropouts you get while loading new files; you can upsample a playing [readsf~] and record that into your table. If you're trying to load 5 stems at a time, you will have to lower the rate accordingly, but here is the basic idea, in my post 3 or 4 posts down in this thread:

http://puredata.hurleur.com/sujet-7578-accuracy-limitation-soundfiler-dropout-potential-fix

otherwise, yeah,,,,get more RAM, or make sure your RAM is all assigned properly to max it out for pd's use. Pd uses 32 floats too, so i THINK that your tables will be saved full of 32 bit floats - which significantly raises the amount of RAM used by each of your stems.

I've had a look at that patch in the other thread, but to be totally honest. it's really confusing me. Do you mean that I need to load the stems into [readsf~] and then re-record each one into [tabread4]?
Would I have to do this each time I load the project? Is that not going to take hours to re-record 45-50 audio files?

• how would I go about making sure that all of my RAM is assigned to just PD (on Windows 7)?

Thank you so much for your help. I've a week to try and get this sorted.

I've 6gb of RAM (which should really be more than enough) and 1.65GHz processor. Will upping the Page Filing work?

EDIT: Nope, increasing the Page File doesn't do anything. I crash after about 4 tracks (20 files) are loaded in.

you load the stems into [reads~] and then play them back MUCH faster than realtime, by using upsampling. I'd hazard a guess that you could get 4 or 5 stems into tables that way in less than 1 munute (depending on how long your stems are, etc)

so, my suggestion is to load your stems for the next track in the background while your current track is playing - that way you don't need to have all 45 stems loaded at once. You should be able to just have your current track's stems loaded, and then your next track's stems loaded - for a total of no more than 10 stems in memory at any moment.

i'll do you up a better demo patch, hold tight...

here you go... i have tried to explain as much as possible, but obviously you will need to fill in the blanks for yourself.

what you need to do to make that patch work with your system:

devise a way to make them stems load alternately from tables 1-5 and then 6-10 and then 1-5, and then 6-10 etc,,,

it's as simple as resetting the counter at the top to 1 every time after it reaches 10.

http://www.pdpatchrepo.info/hurleur/upsampled-loading-stems.pd

Thank you so much for the patch. I've managed to get it sort of working.

I can load the tracks into the tables, but I'm not sure where to go next.
I've used this video

as a guide into making the start, stop, ff, rw and pause controls. In the video that relies on [soundfiler] which confuses me a little in this instance as I've no idea how I would implement the controls in the above panel. It seems that it uses the incoming information from [soundfiler] to control it all which I can't figure out how to do in this patch.
I can't actually figure out how to get the tables to play (I'm still quite a novice with PD) I'm going to keep trying though.

I cant seem to get any output with [tabplay~ $0-stem-1] either.

I also keep getting "dac~: bad vector size" in the panel too. I'm so lost

Thank you

the output of soundfiler is just the sample length, in bits or bytes or something (quite embarrassed that i really don't understand basic computer stuff like that)

the bad vector size error is probably because you are trying to load and play files in the same subpatch or abstraction. You need to make one upsampled subpatch for loading the files, and a different one with the default samplerate for playback.
for the upsampled patch, you need to use the [switch~] or [block~] object like in the patch i sent you, but for default samplerate you don't have to use any objects. Just make sure you keep the upsampled stuff and the playback stuff in different subpatches and it should be ok.

as long as you know what questions to ask, we're here to help you

Mod.... You're an absolute star and a wizard.
I think I've managed to get it working. Although is there a way to get the samplerate to then be used with tabread4~? Or is there a way to do start, stop, ff, rw and pause with tabplay?

I think once I know how to do that I'll be away (he says now with his fingers firmly crossed)

I've managed to get everything else that's needed done like the channel selection and routings sorted.

I'll upload the final patch once it's done for others to play with.

Yup so I've got it all working finally.

Still struggling with getting the sample rate and stuff to be able to use tabread4. Is there a way to either get the samplerate without using soundfiler? Or with using soundfiler in the above method.

Or is there another way to ff or rw with tabplay?
I've managed to get pause and play, now really just rewind and fastforward would have everything absolutely perfect

what do you mean by ff and rw? do you want to playback in reverse, and at fast speed, or would it be enough just to jump to a new location in the file? tabplay~ will let you jump from place to place by accepting an offset value in its inlet. But for reverse, or speed-changed playback, that's what you will need tabread4~ for.

unless you want to loop though, you might be fine with just setting up an 'open-ended' tape style, so that reverse just keeps reversing forever and ffwd just keeps going forward forever. Not sure if that's too clear, but i'll post a patch for you to see, hang on...

this patch uses a normal soundfiler to load the sound, but can be easily modded to use the upsampled background loader too.

note that the output of soundfiler is not used at all in this case, anyway.

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

Sorry, rw= rewind + ff = fast forward.

I figure if all the stems are upsampled and loaded into tables, then I can use tabread4 to play them, and manipulate the phasor to fast forward or rewind (I did that in my old patch using the tutorial I posted before). So I guess playback in reverse, but no need to loop
I just need to figure out how to get the samplerate information when using readsf or tabwrite in the initial upasmpled patch so that I can then use tabwrite4 to play and ff and rw and pause etc

are you sure 'samplerate' is what you are after? if i'm not very mistaken, then the samplerate for playback will always be the pd default of 44100hz unless you change that.

so, when you load the sounds with readsf~. you are playing back at 'normal' speed, multiplied by your upsample rate. However, tabwrite~ is also recording at the upsampled rate. So, when you playback with tabplay~ or tabread4~ or whatever, then you should be playing back at default samplerate of 44100hz.

are you sure you don't mean sampleLENGTH? that's what is output by soundfiler. If you want to get that, you can either use an external like [wavinfo] (not sure how that goes with dropouts though) OR, you can get a quite approximate length by banging a [realtime] object when you start playback of your file with readsf~, and stop the timer when readsf~ outputs its bang that it finished playing. That would give you the time in milliseconds, which you could quite easily convert to samples (44.1s/ms at default samplerate)

but, just have a look at the patch i just posted before...it does playback without phasor, and uses a [vline~] instead, but in my opinion works better than the phasor~ approach and is more flexible.

Well; I've finally got it all working smoothly
Mod you're a life saver and I really appreciate your help.

I did mean samplelenght I was being silly. The last patch confused me at first but I've finally got it all figured out.
I leave in about 48 hours to play so I have got it done just in the nick of time with enough time to test it all and ensure that I know what I'm doing

I'd have been so stuck without your help.

I just thought; would I be correct in thinking that the upsampled loader stems patch is writing everything as mono?

If 3 of the 5 stems are stereo would I need to do [tabwrite~ $0-stem-1-left] and [tabwrite~ $0-stem-1-right] with 3 extra tables for the 3 stereo files?

yes that's correct.

also make sure you use [readsf~ 2] to get both L+R channels read.

and of course, you will need extra [tabread4~] objects for playback as well

good luck! let us know how it goes.

Ok so I have

_________________________[6(___[7(___[8(
[pack f $0]
| |
[open $1, 1( ________________ [set $2-stem-$1-left]
| / /
[readsf~ 2]___________/______ /
| / \ /
[tabwrite~ $0-stem-1-left] [tabwrite~ $0-stem-1-right]

Achtung ! You should not use dollars in objects the same way you use them in message boxes !
Use [set arrayname< to bound [tabwrite~] to a specific array.

a bit hard to tell what's going on with the ascii art there, so plz post your patch, or at least the section that's causing the trouble.
as dwan says, it looks like a mixup with $ substitutions.

I managed to sort it all out, got it all working but then the routings all went wrong. I will need to rebuild from scratch but at least now I know what I have to do. It's frustrating and I'm gutted I didn't get it sorted in time, but at least I'll be able to do it next time.