Muse Choir: Women, playback not working, and major performance issues with choir playback

[samsoeder]

Issue type

Muse Sounds bug

Bug description

The Muse Choir women patch does not playback sound at all. And when trying to playback a 32 bar 8 part choir score my PC grinds to a halt, having higher CPU usage than an entire orchestral score with over 200 bars of music.

My PC specs:

Intel i5
NVIDA GeForce GTX 1050 TXI
16 GB ram

Steps to reproduce

Try to playback Choir women patch
no sound

Screenshots/Screen recordings

2024-05-07.13-09-08.mp4

MuseScore Version

OS: Windows 10 Version 2009 or later, Arch.: x86_64, MuseScore Studio version (64-bit): 4.3.0-241231433, revision: github-musescore-musescore-5f36e74

Regression

I don't know

Operating system

OS: Windows 10 Version 2009 or later, Arch.: x86_64,

Additional context

No response

[MarcSabatella]

It’s a known issue that “full choir” (and “women”) are broken above G4. Other sounds should work.

[samsoeder]

Ok, thanks.

[bkunda]

Not sure whose court this will fall into, but seems like it needs fairly urgent attention @RomanPudashkin @matthewreadbass. Thank you!

[MarcSabatella]

See #22572 for the original GitHub report about notes above G4 not playing for Women and Full Choir. Apparently not everyone can reproduce this, but it's definitely the case for at least some of us using Muse Sampler 0.6.1.162 and Muse Choir 0.3.15 on both Linux and Windows (multiple reports from both operating systems).

I would guess the CPU issue is not related, but I have no special insight there.

[samsoeder]

I've done some testing to give some more context

Orchestra score with 30 parts and 295 measures of music: CPU usage between 40% and 70%

Orchestra score with 30 parts and 350 measures of music: CPU usage between 50% and 80%

Choir score with 4 parts and 32 measures of random notes (using the standard SATB score template with the MuseSounds SATB patches: CPU usage between 26% and 38%.

Same Choir score using MuseSounds choir: Women and men patches: CPU usage between 32% and 40%.

Same Choir score using the women and men patches but each part was duplicated to create 8 parts, 32 measures: 44% and 52% with a short spike at 68%

The original choir score in the bug report: First time playing the score I got 50% and 68% CPU usage. But interestingly when I went back to the beginning and played it again the CPU usage shot up quickly to 100%.

I went back to the random note score with 8 parts and played it twice to see if it usage spiked again but it did not.

I did all the test with only a single score opened and the same background processes running. Memory usage stayed stable for all the test.

I can send the project file over if needed. I'm not sure how to attach it to this comment.

Edit: I have done some further testing and I have recreated the performance issues in the random note score. I added the new staff text feature to change the articulation every 4 bars and saw a slight CPU usage increase. However when I added dynamics the CPU usage started going to 100%. So it seems the issue might be with loading the articulation samples?

2024-05-10.11-03-30.mp4

[samsoeder]

Ok final test.

I closed musescore and reopened it. I opened the random note score and played the 32 bars 4 times. Each time eyeballing the average usage stats at the beginning and end of each test.

Before opening score:
Cpu: 8%, mem: 30%

test 1:
start: CPU 27%, MEM: 33%
End: CPU 40%, MEM 35%

test 2:
start: CPU 45%, MEM: 37%
End: CPU 58%, MEM 41%

test 3:
start: CPU 74%, MEM: 47%
End: CPU 82%, MEM 52%

test 4:
start: CPU 94%, MEM: 57%
End: CPU 100%, MEM 63%

It seems something is getting stuck and causing the usage to go up each time the score is played.

[MarcSabatella]

The range issue appears to be fixed in Muse Choir 0.3.17. Again, I have no insight into any CPU issue.