Anukari 3D Physics Synthesizer: Beta Released

[pekbro]

Interestingly, after updating to win 11 24H2 performance with this on my Core i5 and integrated graphics is greatly improved. A setting to scale and fix the window size decrementally would be nice...

[GPU AUDIO]

AUTO-ADMIN: Non-MP3, WAV, OGG, SoundCloud, YouTube, Vimeo, Twitter and Facebook links in this post have been protected automatically. Once the member reaches 5 posts the links will function as normal.

I tried on my Mac M4 Max. 96 kHz, 96 samples - drop outs for a minimalistic setup, 1 stand-alone instance.

What's the way to run it multiple instances to avoid any drop outs, is there any confirmed configuration for this available? I also have 5090 Windows setup and can run things there to try on.

Also, switching project settings when instantiated causes crash (I changed 48 kHz to 96 kHz) with plugin loaded in Logic Pro, here is the dump: https://paste-bin.xyz/8158979 (https://paste-bin.xyz/8158979)

[GPU AUDIO]

I played around with SpiderWeb for a bit, and it's awesome!

It uses roughly half of my M4 Max's GPU resources (128 samples at 96 kHz) in Logic Pro, which is quite heavy per single instance. I haven't tested it with automation curves yet to see if performance drops after 3 or 4 instances, or if it scales differently (not sure what the real device saturation is).

As the author mentioned, the device-side processing is already extremely fast, so the bottleneck seems to be the CPU<>GPU data transfer, along with optimizing kernel calls (fusing/coupling) and properly overlapping commands.

That said, I see it more as a sound design tool rather than something requiring multiple instances, it's not a synthesizer in that regard.

UPD: activity monitor says it's 80% per SpiderWeb instance on my M4 Max system when it comes to the system-level GPU usage. I hope these are just persistent threads + dummy kernel ran in the loop where you run multiple instances or implemented any other smart approaching managing multiple instances and resources

UPD2: yeap, I open a secondary instance which is stand-alone, and it killed everything with drop-outs, so there is no smart way of managing kernels, the second instance took whatever left from the device and the first one caused the drop-outs despite just leaving in the background, untouched

[anukari-music]

For folks who are unaware, the "GPU AUDIO" account that posted the last couple of replies is a software vendor that, as their name suggests, supports audio plugin manufacturers in making use of the GPU for audio processing. I've met the CEO, Alexander, who I assume is posting, in person and he's a nice guy. To summarize the situation, he really wants Anukari to license his company's technology, and I think he truly believes it will help with any performance issues. However, Alexander can be perhaps a bit "overzealous" in his product outreach, shall we say. 3 out of 4 of all the posts he has made on the KVR forums are on this thread. I've explained to Alexander a couple of times why his product's not a great fit for Anukari, but I guess the thinks that I don't know what I'm talking about. That's fine, maybe I don't. This thread ain't the place to discuss that, though.

Alexander: I am not going to reply to you here any further. Please feel free to email me if you want to have a serious discussion.

Anyway, I do have some early updates regarding performance on Apple hardware, particularly the more powerful units including their "Max" level hardware.

A couple of days ago I wrote an open letter to Apple: https://anukari.com/blog/devlog/an-appeal-to-apple

It got a lot of publicity and the right folks at Apple reached out to me. We had a super productive conversation. Long term I will be working closely with them to help with their work to improve support for weird software like Anukari. In the short term... they opened up about some of their hardware details under NDA, and gave me some hints. So I'm quite optimistic that improvements are forthcoming. I better get to work!

[pekbro]

Make sure to hold out for the big bucks if apple tries to buy you out. Heh

Nice to be popular I guess.
Mostly anyway…

[Funky40]

Anyway, I do have some early updates regarding performance on Apple hardware, particularly the more powerful units including their "Max" level hardware.

A couple of days ago I wrote an open letter to Apple: https://anukari.com/blog/devlog/an-appeal-to-apple

It got a lot of publicity and the right folks at Apple reached out to me. We had a super productive conversation. Long term I will be working closely with them to help with their work to improve support for weird software like Anukari. In the short term... they opened up about some of their hardware details under NDA, and gave me some hints. So I'm quite optimistic that improvements are forthcoming. I better get to work!

ahh, that´s great to hear. I´m on a M4-Max.

i´d like to make this point please:
i use GP (GigPerformer) as my plugin-Host.
it is in my opinion the best live play host out there.
Quasi the whole GP community would much likely agree with me. (well, no doubt on that)
That´s at leats true for handplaying, non sequenzing, and if auto-modulations don´t play a role, or at least not a big one either.

would be great if you @Anukari would be willing to add GP to your testing platforms/ Hosts, in case you test on some.
Maybe reach out to David. Who knows, maybe they´d grant you a NFR license ?

[anukari-music]

ahh, that´s great to hear. I´m on a M4-Max.

i´d like to make this point please:
i use GP (GigPerformer) as my plugin-Host.
it is in my opinion the best live play host out there.
Quasi the whole GP community would much likely agree with me. (well, no doubt on that)
That´s at leats true for handplaying, non sequenzing, and if auto-modulations don´t play a role, or at least not a big one either.

would be great if you @Anukari would be willing to add GP to your testing platforms/ Hosts, in case you test on some.
Maybe reach out to David. Who knows, maybe they´d grant you a NFR license ?

Actually, David already reached out to me and we have exchanged NFR licenses! I am not sure which one of us will have time to test it first, but it is definitely going to happen. (I am prioritizing the performance work at the moment, but after that.)

I've never used Gig Performer so I'm excited to try it out a bit.

[Funky40]

Actually, David already reached out to me and we have exchanged NFR licenses! I am not sure which one of us will have time to test it first, but it is definitely going to happen. (I am prioritizing the performance work at the moment, but after that.)

I've never used Gig Performer so I'm excited to try it out a bit.

Nice one !

[buffalo roam]

Suggestion: I'm wondering if most of the initial presets are overkill in programming. Visually striking, for sure, but maybe more complex than they should be, and therefore over-demanding on computer capabilities. Maybe developing a bunch of less complicated presets would both help out users with computer limitations while opening the door more to individual experiments in tweaking them.

[v1md]

Suggestion: I'm wondering if most of the initial presets are overkill in programming. Visually striking, for sure, but maybe more complex than they should be, and therefore over-demanding on computer capabilities. Maybe developing a bunch of less complicated presets would both help out users with computer limitations while opening the door more to individual experiments in tweaking them.

The default instrument seems to be fairly simple in design, one mass and two springs. Still, I'm struggling to get two instances of Anukari open. Not great ...

[tonycore]

I’m also a Gig Performer user, very happy to hear that you and David are talking.

Been on the road/out of the country, will be purchasing the beta after I get back tomorrow. Looking forward to this!

Tony

[anukari-music]

A couple of updates for folks following this thread:

1. For folks with powerful macOS machines who tried Anukari and it didn't perform well, I'm working with Apple to fix this and the first release (0.9.5) with improvements is available at https://anukari.com/download. I'd love to hear feedback on whether this helps. More improvements will be on the way after this release. (Note: you still need a buffer size of 512+.)

2. The first batch of short and to-the-point video tutorials is available. There are a bunch more coming to cover the remaining features, stay tuned: https://www.youtube.com/playlist?list=P ... PJ1853Fybs

[Ah_Dziz]

On top of being very nifty, this thing reminds me of rendering floating orbs in corel in windows 95..... In a good way.

[nanostream]

1. For folks with powerful macOS machines who tried Anukari and it didn't perform well, I'm working with Apple to fix this and the first release (0.9.5) with improvements is available at https://anukari.com/download. I'd love to hear feedback on whether this helps. More improvements will be on the way after this release. (Note: you still need a buffer size of 512+.)

I’ve played around briefly with the latest beta last weekend, and performance was indeed noticeable better on my M4 Pro. But it still felt a bit clunky, and after reading your blog post about your conversation with Apple I decided to wait for the next update.

I really hope you/Apple manage to bring down latency somehow. Needing a 512 samples buffer size is a bitter pill to swallow, especially with MPE.

[WasteLand]

1. For folks with powerful macOS machines who tried Anukari and it didn't perform well, I'm working with Apple to fix this and the first release (0.9.5) with improvements is available at https://anukari.com/download. I'd love to hear feedback on whether this helps. More improvements will be on the way after this release. (Note: you still need a buffer size of 512+.)

I’ve played around briefly with the latest beta last weekend, and performance was indeed noticeable better on my M4 Pro. But it still felt a bit clunky, and after reading your blog post about your conversation with Apple I decided to wait for the next update.

I really hope you/Apple manage to bring down latency somehow. Needing a 512 samples buffer size is a bitter pill to swallow, especially with MPE.

a 512 sample buffer size, can have for instance a 12ms output latency, for the audio interface.

which in my book, with MPE, is no problem.

normally i do 256 sample buffer size, which is around 6.3 ms output latency.

i can't say it is a bitter pill, i work with it, don't notice it...

you can adapt (or not?).... latency sensitivity is personal, and what you are used to.