My DIY Multi-Speaker / Multi-Sub auto optimizer

[umamiaudio]

Anyone up to try my home brew speaker calibration system? Lol. It's closed source. I'm giving away a free licenses to every who will take time to help me beta test it. You need your speakers to be connected to a computer and right now the calibration files can only be loaded as a VST/AU/AAX. You will also need a cal mic. The algorithm that does the actual optimization took me over two years to perfect. I swear it's better than anything on the market. My inspiration for this project was basically "I think I can do better than this". It supports multi-subwoofer systems and will integrate the whole system for you automagically and there's nothing for you to set at all (almost). It looks at everything... time domain response, your distortion response, room modes, etc.

https://umamiaudio.com

If you're using MSO, give this a spin! Also for those asking, is DOES NOT try to maximize your bass SPL because that usually comes at the price of sound quality. It DOES make the response as linear and as tight and defined as possible at every seat while sacrificing as little volume as possible (because it also accounts for distortion). I'm here to answer any questions. Very open to questions and criticism. If you try it and find that it doesn't work well for your system I will personally have a look at your measurements and figure out exactly what went wrong and fix it for everyone.

 

[anlin]

I took a look at your website and it's a very interesting product. Not many MIMO room correction solutions out there aside from ART, so taking on a project like this as a solo developer is commendable. The pricing, however, seems excessive for an unproven product that has just launched. Having to purchase a subscription to maintain calibration access on top of a $1000 "perpetual" license especially doesn't sit right with me.
What kind of licenses are you giving out to beta testers? I would be willing to give this a spin if you can offer a true perpetual license that doesn't eventually lock me out of the software or calibration ability. I'm already running DLBC as a VST on a dedicated, Dante-enabled DSP server, so your room correction tool would fit right into my current setup..

 

[umamiaudio]

The beta testers who help me to actually test it and put it through it's paces get a perpetual license. You can keep using any calibrations that you make within the first year perpetually in this case. The software updates are always free for everyone.

After the first year you do need to purchase one of the calibration subscriptions if you want to create new calibrations, but you can keep using existing ones forever. This is done for serval reasons:
1) most users won't have the hardware resources required to generate these calibrations as they are heavy and run on GPU resources
2) it acts as copy protection for my code
3) I have real running costs for these servers
4) the optimization algorithm itself has a lot of dependencies. This would make the install size ~20GB instead of a small download and make it less reliable and maintainable as a solo dev since all of it would need to run reliably on user-end environment.
5) You also support the UI/UX updates for the plugins and wizard, which are always free for everyone.

However there are benefits to this as well:
1) The idea is that the algorithm should be improved over time and you get whatever is the latest and greatest I have to offer at that time for a tiny cost. I already have many different ideas for improvements to both the user experience and DSP but for now the focus will be bug fixes and fixing any edge case issues that come up.
2) The measurements are stored server side and I can have a look at individual measurements and calibration results to give users individualized support if needed, as well as use them to fix edge cases and further improve the algorithm.



You say it's untested but you can very easily test it for yourself and even compare it to your current system head to head. The beta is totally free and gives you 3 months of access. This could be extended or closed, depending on if I can get enough people to try it and have it running without any issues.

All in all, I think the pricing is fair and still cheaper than the competition if you pay for 10 years of calibration up front (which is actually kind of unrealistic because most people are not going to change their setup every year)

 

[anlin]

A couple more questions:
1) What kind of filtering scheme does your software use, minimum/mixed/linear phase?
2) What kind of hardware resources are needed for running the processor VST?
3) What kind of latency numbers would we be looking at for multi channel, multi sub, full range correction?
4) Any more specifics you can reveal about the calibration process? The large download size and GPU utilization leads me to believe it is based on some kind of an AI model.

As for licensing, it's your software and you're free to license/price it however you feel like. While the product itself is very interesting, I'm personally dissuaded from using it for a few reasons:
1) I don't want my room correction software to be SAAS
2) Even if you have a perpetual license option, it still requires network connectivity. What if you company goes under or your servers are down for whatever reason? Investing $1000+ in an always-online product by a fairly anonymous, previously unknown company gives me pause
3) It's not really appropriate to compare the pricing to ART or Trinnov directly, as both of the above come with hardware while your solution requires the user to bring their own. The closest (and only) competitor I can think of is the DLBC VST at $700 (with unlimited calibration access!), although it is not MIMO.
4) I personally can't agree to an additional, fairly high per-calibration fee (because like you said, most people only recalibrate when upgrading equipment or moving, so less than 1x/year, necessitating a payment each time).

I just thought I'd give you my .2c since you asked for questions and criticism. That being said, good luck with your product, the market really could use another option.

 

[ShadowBoxer]

Are there any installation alternatives other than downloading and executing an unsigned [exe] with super-user permissions? (which triggers typical security warnings and there are no listed CRC hash values on the download page).

 

[ShadowBoxer]

Anyone else having trouble with the Discord channel?

 

[dziemian]

I am running stereo plus 6 subwoofers using Jriver vst plugins. I beta tested Dirac DLBC and played with MSO as well. I am interested in beta testing your software? What do you need from me?

 

[ShadowBoxer]

What do you need from me?

You can sign up for Beta testing at the website.

 

[umamiaudio]

A couple more questions:
1) What kind of filtering scheme does your software use, minimum/mixed/linear phase?
2) What kind of hardware resources are needed for running the processor VST?
3) What kind of latency numbers would we be looking at for multi channel, multi sub, full range correction?
4) Any more specifics you can reveal about the calibration process? The large download size and GPU utilization leads me to believe it is based on some kind of an AI model.

As for licensing, it's your software and you're free to license/price it however you feel like. While the product itself is very interesting, I'm personally dissuaded from using it for a few reasons:
1) I don't want my room correction software to be SAAS
2) Even if you have a perpetual license option, it still requires network connectivity. What if you company goes under or your servers are down for whatever reason? Investing $1000+ in an always-online product by a fairly anonymous, previously unknown company gives me pause
3) It's not really appropriate to compare the pricing to ART or Trinnov directly, as both of the above come with hardware while your solution requires the user to bring their own. The closest (and only) competitor I can think of is the DLBC VST at $700 (with unlimited calibration access!), although it is not MIMO.
4) I personally can't agree to an additional, fairly high per-calibration fee (because like you said, most people only recalibrate when upgrading equipment or moving, so less than 1x/year, necessitating a payment each time).

I just thought I'd give you my .2c since you asked for questions and criticism. That being said, good luck with your product, the market really could use another option.

1) Mixed
2) It should run on any regular computer. It uses the very efficient multi-threaded FFT convolution with SSE optimizations
3) <10ms -- less than 1 frame at 60fps. There are plans to reduce this number but I need more measurements first.
3) What's the price of the most entry ART / Trinnov hardware? Mine will run on something cheap like a scarlett which is only a few hundred dollars.
4) I am a bit hesitant to reveal any specifics because if this becomes popular, companies like Dirac have virtually unlimited resources and I wouldn't want them to know where to even start looking. While it's non-trivial, it's the little creative bits of problem solving and the philosophy I've put into the algorithm that makes it special.

1) Yeah I get that, but unfortunately this is unavoidable for now for the reasons I have explained earlier.
2) Network connectivity is also used for copy protection. There is a 1 week grace period on the licenses where it will keep working even with no network or the server is down. If the company goes down I'll probably release it as open source or something. I'm here to do good and make some money while doing it, not to take advantage of people. If you are unsure about investing $1000 upfront you can always start on a subscription plan. Or even better, try it now for free for 3 months!
3) But how much is their hardware? Depending on your channel count needs, you can pick up a Scarlett or something similar for a few hundred dollars.
4) It's not per calibration, it's per month / year. I think if you're running a 9.4.4 system or something that requires the highest-tier license, you should be able to afford $49 once every few years.

 

[umamiaudio]

Are there any installation alternatives other than downloading and executing an unsigned [exe] with super-user permissions? (which triggers typical security warnings and there are no listed CRC hash values on the download page).

No but you can try to upload it to virus total. Currently there seems to be one false positive. I'm here to build reputation and make money in the long run not spread some viruses, otherwise I would just say all of it is free or something and have some bogus DSP calibration to try to get you to run the exe.

The mac executable is signed and notarized by apple, but for the .exe I need an OV cert which requires in an person notarization with a third party and I am traveling at the moment. I'm hoping to be able to get this done by end of June.

Posting CRC hash is useless because if my server is compromised then the attacker could just change the CRC value on the page as well, but actually the website is hosted by github and it's very secure.

Is discord working for you now? I can see that you have joined on my end.

 

[ShadowBoxer]

The mac executable is signed and notarized by apple, but for the .exe I need an OV cert which requires in an person notarization with a third party and I am traveling at the moment. I'm hoping to be able to get this done by end of June.

Is there some reason a zipped VST is not possible? Executing an unsigned binary with full privileges is a security non-starter for me. Installing a VST plug-in in regular user space is much less risky.

 

[ShadowBoxer]

Is discord working for you now? I can see that you have joined on my end.

I verify email...discord acks and says verified...and then reverts back to please verify your email. I haven't been able to break the loop and discord says "failed to load messages".

 

[umamiaudio]

Is there some reason a zipped VST is not possible? Executing an unsigned binary with full privileges is a security non-starter for me. Installing a VST plug-in in regular user space is much less risky.

Absolutely! I made these updates just for you so please actually try it. You'll need to install these dependencies:

https://go.microsoft.com/fwlink/p/?LinkId=2124703

https://aka.ms/vs/17/release/vc_redist.x64.exe

Then you can download the wizard from:

https://umamiaudio.com/downloads/Calibration%20Wizard.exe

And the VST from:

https://umamiaudio.com/downloads/vst.zip

put the whole VST folder (Umami Calibration Loader.vst3) where your DAW normally scans for VST
Our default install goes to C:\Program Files\Common Files\VST3

The wizard should go under C:\Program Files\Umami Audio\bin

The wizard doesn't NEED to be there but some functions that open the wizard from the VST will not work unless it's placed there.

 

[umamiaudio]

I verify email...discord acks and says verified...and then reverts back to please verify your email. I haven't been able to break the loop and discord says "failed to load messages".

Yeah, discord can be kind of buggy sometimes with signups. Unfortunately I have no control over this. I would either sign up with a different email or just wait a bit and try again.

 

[john61ct]

Can this be used on a standalone pass-through basis? I rarely play music from a PC.

Maybe Raspberry Pi as a convolver?

 

[umamiaudio]

We might support Linux at some point, but not right now. Basically whatever you use it with needs to be able to run win/mac VST or AU. I'm also not sure how much CPU power a Raspberry Pi actually has.

 

[ShadowBoxer]

fyi, there's a typo in the VST download url above (I worked around it).

 

[umamiaudio]

fyi, there's a typo in the VST download url above (I worked around it).

fixed, thank you for letting me know

 

[Ktacos]

As a mix engineer, this looks right up my alley. I'll give it a shot.

 

[ShadowBoxer]

~ Got the basic P-Soup wizard working.
~ Got P-Soup to recognize/test the Umik-1 (some possible gain issues, but I deferred that bc of below)
~ Having trouble getting P-Soup to work nicely with the HW sound drivers.

ie I normally use WASAPI (successfully). However, P-Soup only enumerates ASIO for reasons I'm not clear on? And ASIO doesn't provide sound output on my HW for underlying reasons TBD.

For example, using my (stereo) HW with my other working configurations:

So, I'm at a dead-end due to these underlying IO issues.