TX16Wx 3.2.2 beta available

Official support for: tx16wx.com
RELATED
PRODUCTS

Post

New build up. Now round-robins voices across threads more, as well as max slot poly == max global poly and some small perf tweaks. Also better oscillator handling in saved materials (not really visible, but...).
TX16Wx Software Sampler:
http://www.tx16wx.com/

Post

Thanks again for the continuous refinements, really! I still don´t notice a major improvement in terms of CPU efficiency. With an I7 3770k, from 2013, I see this numbers while loading the same sfz - I won´t name the library again, don´t worry ;) - :

Sforzando : aprox. 1,30 %
TX16WX : aprox. 7 %

There was a slight gain lately with these tweaks; maybe Tx16 was closer to 8-9%. Of course, this could be due to the sfz script in itself, as you suggested (I haven´t had the time to build the instrument from scratch with the samples). I simply comment this to raise the bar and see how much can Tx16wx gain in terms of CPU usage. If you say that there´s a limit regarding the possibilities of TX16 in this department, I obviously won´t discuss the topic anymore - I´m really more on the composition/performance side of things, and the technical aspects of all of this tend to be rather abstract to me -. For things that don´t need sample tweaking, I can always use Sforzando, of course. I´m already putting to good use Tx16 and testing its capabilities on real projects, so I´m more than happy with it.

Oh. one thing : I still can´t modify the "Max voices" box...

Thanks for listening, Calle. One way or the other, Tx16 is better and better as years pass by.

Post

Are you using TX16Wx Pro? Multi core processing and variable voice count is only available in Pro mode. Similarly, you need to set the number of cores used. You can enable Pro trial by simply clicking "use pro features". You'll get some nags, but not many the first week. :-)

I doubt TX will ever be able to match a static SFZ player in raw performance. The general purpose playback engine comes with some inherent costs. Plus I am also somewhat lazy/protective of my sanity, so I'd like to keep some of the code path reasonably cleanly organized. :-)
TX16Wx Software Sampler:
http://www.tx16wx.com/

Post

I´ve tried the Pro features. As I´ve been using for some time the free version, the trial takes that into account and I get a long countdown before I can use it (I continuously get TX16 to crash, by the way, when I click on "Use the free version"). With the Pro features, I see the CPU usage improves considerably indeed (from 6-7% to 2%, approximately), but with 128 max voices, voice-stealing is quite exaggerated. In a double trill passage, as in the test project I sent on another thread, way many voices are left out. And if I increase the max voices number, the CPU goes to the moon. With something like Piano In 162 - you won´t dare to tell me that it sounds like a 90´s Akai Sampler, will you? :) -, things get considerably worse.

Post

One thing : The Countdown windows say I´ve used Pro 170 times, but that is not true, as far as I know. I installed TX16 free some time ago (apart from using it some years ago), but I wasn´t using Pro until the other day.

I get a nasty situation with 3.2.2 : I can´t return to the free version without getting an unhandled exception error, which itself causes Reaper to crash repeatedly. It doesn´t happen with 3.2.1d.

Post

I never said the piano banks you've mentioned sounded like 90:s Akai. I said the 90's Akai programs sound better. Because they do, because they are sampled by very good sound engineers, with measurable signal-to-noise ratio, proper post processing, and lo, reasonable file sizes. ;-)

The "Piano In 162" nowhere as terrible as the salamander, but still suffers from the insistence on near-minus-90db signal levels and literally seconds of ambient silence in the samples. So incredibly wasteful. Don't get me wrong, the sound is pretty nice, but could have been made so much smaller and performant. And should have been miked way hotter.

One good thing about it though, it is _not_ using unnecessary layers, but instead some pretty neat key switching. That TX unfortunately reads incorrectly, because the locc/hicc opcodes are not supported. Until next build, where they are. Now (with bug fix) the piano plays at one sample per key pressed. And handles the pedal as expected. Nice.

On that note I would like to point out that a piano has an absolute polyphony of 88. No more, sometimes less. Any "ambient" post-string-sound is _not_ the string, but box echo, which should (imho) not be represented using samples, but a short box reverb. But hey...
While I am ranting: piano strings are about as monophonic as they come. Sampling them in stereo instead of using a key-to-pan modifier + reverb is (again imho) a crime against proper studio nature. :ud:

Now to bugs:
- I found a regression in the multi-core renderer that made it loose notes (or rather not copy them to main output). Fixed in next build.
- I also tweaked the disk thread some so I got some better performance. Doing some massive key mashing I get pretty neat cpu numbers now. They are still in mid-upper single digit range, mainly due to the samples in the bank being FLAC (costs some processing power to stream)
- Also fixed the disable-pro-and-multi crash.
- And tweaked the nag code to be a little less zealous in counting days. (It used wall clock, not usage clock).

New build will be up later tonight.
TX16Wx Software Sampler:
http://www.tx16wx.com/

Post

Builds updated
TX16Wx Software Sampler:
http://www.tx16wx.com/

Post

Well, this is some fantastic feedback, I have to say! Don´t take my comments on free pianos too seriously, really :hihi: ; your rants just make laugh - and I admit you have some strong points in your observations -. I have to agree with your comments on the waste of resources; after all I´m all for good and lightweight sampled instruments -.

Talking about samples, I´m loading right now (with 3.2.1d) one of the kick samples from the Free Yamaha 9000 kit from Pettinhouse (the one labeled as "Cassa debole.wav") : http://www.pettinhouse.com/html/download.html . Two unexpected things happen :

- Being a wav file, it defaults to some kind of loop in Normal mode. It forces me to set it to One Shot. Needless to say, it doesn´t happen with other samples.

- Also, there are some delayed clicks when you play something. For example, you play two eight notes and afterwards you hear some delayed eight notes clicks.

Immense thanks for listening and improving in a light-speed fashion!

Post

elcallio wrote: Mon Dec 02, 2019 8:26 pm Builds updated
Thanks! I hadn´t seen that post. Is that the normal or beta download? I don´t see any build with today´s date...

Post

Beta build. The dates on the web page don't update (I am lazy). Check the build number in the installer. Also first post in thread. :-)
TX16Wx Software Sampler:
http://www.tx16wx.com/

Post

SDG wrote: Mon Dec 02, 2019 9:25 pm Well, this is some fantastic feedback, I have to say! Don´t take my comments on free pianos too seriously, really :hihi: ; your rants just make laugh - and I admit you have some strong points in your observations -. I have to agree with your comments on the waste of resources; after all I´m all for good and lightweight sampled instruments -.

Talking about samples, I´m loading right now (with 3.2.1d) one of the kick samples from the Free Yamaha 9000 kit from Pettinhouse (the one labeled as "Cassa debole.wav") : http://www.pettinhouse.com/html/download.html . Two unexpected things happen :

- Being a wav file, it defaults to some kind of loop in Normal mode. It forces me to set it to One Shot. Needless to say, it doesn´t happen with other samples.

- Also, there are some delayed clicks when you play something. For example, you play two eight notes and afterwards you hear some delayed eight notes clicks.

Immense thanks for listening and improving in a light-speed fashion!
Yeah, the samples have loops (and roots) in them. Not TX:s fault. If you map a sample with loop(s), TX will assume you are mapping a sustained instrument. Just turn off loops for the region(s) (sustain and release) instead of using one-shot. You might want to override root key to "no root" (delete in edit box).

Not sure what "delayed clicks" you are talking about. The samples play back fine for me?
TX16Wx Software Sampler:
http://www.tx16wx.com/

Post

Hmm, in a couple of hours I'll test it following your advice and if it still clicks, I'll send you an audio example...

Post

elcallio wrote: Mon Dec 02, 2019 10:18 pmNot sure what "delayed clicks" you are talking about. The samples play back fine for me?
Well, I sent you a sample audio file to info@cwitec.com for you to listen. Anyway, when I set to 0 the Decay 2 and Sustain, the weird clicks dissapear.

I´ve used the Pro features now. I have to say that CPU usage is way better now, and note stealing has also been fixed. Now it´s much closer to Sforzando with the double trill example : it approximately doubles it (ca. 3% of TX against 1.5% of Sforzando), which is reasonable, taking into account the difference you mentioned between an static and multi-purpose engine. Kontakt, for example, is closer to Sforzando - ca. 2,3%, and always talking about Piano In 162 -, but that must be related to some multi-core manipulation. Anyway, the improvement in TX´s performance is quite welcome.

Now, regarding the Pro trial, I think it has shifted from severe into plain punishing. The first time I turned it on it started at 96 seconds or so, more or less as it happened repeatedly in the previous build. I turned off the plugin and when I activated it again it jumped to 190 or so seconds. The third time it went straight to 560 or so seconds (!!!). After that, I wouldn´t be able to test any Pro feature again, I have to admit. It´s too unfriendly, maybe because it still takes into account the free usage time.

Post

Some things I also observe while working with TX :

- I´m getting an "Invalid vector <T> subscript" error sign when I load a bunch of kick samples on different groups and I turn off/on in Reaper.

- Although I set a specific folder to store the reapeak files, it still creates them in the sample folders.

- Choosing the "New program slot" option duplicates an already existing program slot. From then on, any edit is duplicated in both slots, no matter what you do.

Post

SDG wrote: Tue Dec 03, 2019 12:54 pm Some things I also observe while working with TX :

- I´m getting an "Invalid vector <T> subscript" error sign when I load a bunch of kick samples on different groups and I turn off/on in Reaper.
What does "I turn off/on in Reaper" mean? Please be more descriptive/provide exact steps to reproduce issues.
SDG wrote: Tue Dec 03, 2019 12:54 pm - Although I set a specific folder to store the reapeak files, it still creates them in the sample folders.
You need to also set the "Place wave peak file in alternate cache path" to "Always" to force usage of reapeaks path.
SDG wrote: Tue Dec 03, 2019 12:54 pm - Choosing the "New program slot" option duplicates an already existing program slot. From then on, any edit is duplicated in both slots, no matter what you do.
No, you are creating a new slot, but it will still map the same program. Note that changing the slot volume, which is a part of the slot, is not mirrored. Programs a distinct entities that can be mapped more than once. Use duplicate program in the new slot if you want a new program instance.
TX16Wx Software Sampler:
http://www.tx16wx.com/

Post Reply

Return to “CWITEC”