@KMETE If you give all the files a number instead of the name it is easy and you don't need [file glob]

AAaaargh...... should be [declare -path] ... not [declare patch]

Output directly into the inlet of [sf-play2~] that takes the file name.
Then start the player....

If you don't then you need to assign a number to each file.
You cannot add it to the name because [sf-play2~] will only play the actual file name.
You have 3 options.

1. change the names to numbers before you start.... as above.
2. automate replacing the file names with numbers using a command line script after [file glob]....
3. associate, in Pd, a number with the file name........ cart5.zip

For it to run in vanilla you will need to get the missing externals, or replace them with the appropriate [file] objects.
[tof/folderpath] [ext13/strippath] and [folder_list]

Dynamic patching......... there will now be another way using [list store].
It creates a [tracklist_button] for each file in a folder and displays it in the sub-patch [pd tracklist].
Each [tracklist_button has 2 arguments.
The first is the file name.
The second argument..... 1 to n...... is the associated number assigned, and sending that number into [s playit] will send the original file name into [readsf~] which can of course be [sf-play2~] instead.
David.

@fluxmonkey I see nothing wrong, except that the connection to the [delay] leg was made before that to the [pipe] legs, which means the next chord is triggered before the note off messages are sent .
A [t f f] at the last outlet of [unpack] would fix that......
But because the note values are stored sequentially in the [pipe] they are not overwritten by the next chord, so that is not a problem.

Except that if you use something after the sub-patch that restricts your patch to 6 notes..... like [poly 6].... later messaging might be a mess.

In what way is the "timing off"...?
David.

@Dizzy-Dizzy It seems probable that you do not have the correct binary for your system.
Try putting a [declare -lib Gem] object in your patch and saving and re-opening it.
If that still doesn't help then you should try deken (help - find externals) again in case some updates have been coded (but I don't think that will help)..... and then, if it still fails the last resort will be to compile from source (make install) for your system.
At least, when you do that, you should see messages about failures, and if there are only dependancies missing you should be able to get them if they are compatible.
You will need the "latest, bleeding edge" release for that I think.
See the readme here........ https://git.iem.at/pd/Gem

Did you try opening Pd from a terminal after typing

arch -arch arm64e zsh

into the terminal?
David.

@esaruoho I have found the very small investment in TouchOSC (open sound control and midi message capable via wi-fi) very worthwhile........ https://hexler.net/touchosc
It has never let me down on live shows...... since 2011 when I started using it.
I have it on about 20 devices now.

But.... it will only send midi messages via a usb connection...... not osc (which requires a network connection).
It doesn't mind though if you are charging the Ipad over usb at the same time..... it will make the wireless connection.
David.

@Dizzy-Dizzy In the Pd console...... what are the messages just before the "cannot create" error messages for the objects? i.e. does Pd find the objects, or are they incompatible with your system?
I imagine that you still have the entry "-lib Gem" in your file,preferences,startup,startup flags, and the path to the Gem folder set in file,preferences,path.
David.

@bobpell You can also send such a message to the main patch... directly from the same patch....

[;
pd-mainpatch.pd message(

David.

@bobpell First idea......
Put a message to the subpatch into the container patch..... maybe triggered with a [loadbang] or if that doesn't work a [loadbang] + a small [delay 5]...

[;
subpatch vis 0, relocate (width)x(height)+1+1 (x-position)x(y-position)+1+1, vis 1(

See the help file here...... 0.all_msg.pd
Second idea...... not yet....
David.

@60hz It should be as simple as sending 0 or 1 to 192.168.1 4 on port 30000 ...
TCP I think...... the only doc I have found says that it should accept such a connection request.
David.

I seems to me, that in the English language, it writes well.
It's not on the level of D.H.Lawrence or Victor Hugo, but then it doesn't actually "think" and only pretends to have feelings.
It's a robber bot, programmed by the minnions of robber Barons, who will need to be very careful that they do not infringe the property rights of the authors of the original material.
At the moment it is being trained by scraping the web.
The money will be made through data analysis...... for example in medicine..... where governments are keen to give them the non-personalised data...... for a share of the rewards.
It might on occasions accidentally make 2+2=5 and "invent" something.
But as @ddw_music mentioned, "AI" will always suffer, as we all do, from the oldest computer and human virus..... "garbage in garbage out"...... when it interacts with we, the people.
David.

@Dizzy-Dizzy There is a chance that it can work on your M1....... maybe....... https://www.reddit.com/r/puredata/comments/yq36kc/gem_on_a_mac_with_m1_chip/
David.