Feature request: built-in binauraliser for new MIR

[Macker]

Since the majority of people (estimated at about 80%) who listen to music, do so on headphones, it seems somewhat remiss of VSL not to have catered more directly for this huge majority, in the design of new MIR.

Currently, users of new MIR who want to use headphones and also benefit from binaural rendition of MIR's splendid ambisonics spatialisation, are expected to purchase and use a 3rd party external binauraliser plugin.

Recent research in binaural rendering of signals on the ambisonics bus itself (which is not exposed outside MIR), especially the "bilateral ambisonics" technique proposed by Ben-Hur et al [1], promises considerable advantages in definition and quality of ambisonics spatialisation for headphones.

Compared to our only current option of using a binauraliser plugin after MIR's decoding for loudspeakers - a configuration which is unavoidably somewhat of a fudge involving compromised quality - I believe such internal binaural processing would not only extend the attraction and very probably delight of new MIR to many more potential users, but also substantially improve the listening experience of current users who use headphones and have no wish to invest in a full-blown, high quality 'home theatre' loudspeakers-etc setup suited to immersive sound.

Here's hoping such a feature will, in the not too distant future, be added to new MIR.

 

[1] "Binaural Reproduction Based on Bilateral Ambisonics and Ear-Aligned HRTFs". Ben-Hur, Alon, Mehra, Rafaely. IEEE/ACM Transactions on Audio, Speech, and Language Processing, Vol. 29, Jan 2021.

[Dietz]

Thanks for your suggestion, Macker. 

I understand the idea behind it, but I openly admit that we have no plans to add binauralisation ourselves in the foreseeable future. Especially when you consider that MIR 3D is not the actual mix bus in many setups you'll understand that it's the most meaningful way to supply either raw Ambisonics (which MIR 3D delivers up to 3rd Order), or to decode to discrete outputs.

Apart from that: Binauralisation is one of _the_ hot topics in audio right now, and I'm sure that we will see big jumps in market penetration as well as its sonic possibilities. We all look forward to them! :-)

Kind regards,

[Dewdman42]

dearVRmonitor works great for monitoring through headphones!  Also, several DAW's include free binauralizers in them, Cubase and LogicPro with their atmos support both provide.  Might be others as well, those are the only two I'm aware of.  But I think dearVRmonitor sounds a little better then the one built into LogicPro.

The one advantage I can think of for having built in binauralization inside MP3D would be that I think it would substantially reduce the CPU requirement by encoding to binaural before sending back to the DAW...thus drastically reducing the DAW's mixing requirements in terms of audio channels.  My Computer can barely handle 5.1.2 with 100 tracks...7.1.4 was totally out of the question and I believe this was mainly due to all the audio channels coming back into 100 tracks in LogicPro.  If MP3D had option to encode to binaural, then I have a feeling I could use any of the MP3D output formats...and even work with 7.14 and use less CPU.  Just a theory.

This is not as simple as just adding a binaural encoder though because MP3D is designed to pickup the channel count from the DAW's current mixbus channel count.  So things would have to be modified considerably in order to handle the DAW using a stereo bus, but expecting MP3D to process 7.1.4 and encode it to binaural stereo.  

Macker another thing to consider is that binaural is most likely not the format that you want to bounce or distribute.  its just mainly useful for monitoring purposes only.  Well if you are mainly just interested in writing music on you own and not sharing it with anyone anyway, then its a moot point.  But if you plan to distribute it, most likely you would want to distribute Dolby Atmos, which means you would want the full 7.1.4 (or 5.1.2 in my case) but being mixed in you DAW so that you can ultimately generate an ADM file, which ultimate produces a Dolby Atmos MP4, which ultimately is how most people will be able to listen to this kind of content if and when you will be distributing it.  So really...working with the full surround bus in your DAW is the right thing do to....if your computer can handle it...  You can then monitor it from there as binaural if you want or quickly turn off that binaural encoder in order to monitor through real surround speaker config...or to generate ADM from that mixbus using tools in LogicPro.

[Macker]

Dietz, I'd like to add another angle to my suggestion for embedding ambisonics binauralisation in new MIR.

It occurs to me that if DearVR are able to offer their "Ambi Micro" ambisonics binauraliser plugin as a freebie (given of course that it helps Sennheiser's marketing prospects for their AMBEO VR microphone), then surely we're not talking about massively expensive and perhaps risky new technological development - as would be the case for developing MIR to incorporate as an option the Bilateral Ambisonics concept I mentioned above.

Indeed, if I may be so bold, what about a deal between VSL and DearVR to license use of (or relevant chunks of) DearVR's Ambi Micro code, embedded for optional use in MIR?

Just a thought.

[Dewdman42]

dearVRmonitor could be gotten recently at PluginAlliance fro $29.  Might still be available now?  VSL has a running special discount deal already setup with them for, I think $69?

dearVRmonitor also has some licensed sennheiser tech in it,

I can't say enough good things about dearVRmonitor.  It sounds much better then the free binaural encoders built into LogicPro and Cubase, for example.  

[Dietz]

Interesting thread, great suggestions - but as I wrote above: There are no plans to implement a proprietary binaural encoder into MIR 3D in the near future.

[Macker]

Thanks for your response, Dietz.

I guess most MIR users who want to produce top quality spatialised headphone mixes - but without resorting to more laborious and tricky non-orthodox solutions - will get used to having to operate MIR with a recommended 3rd party external quasi-binauraliser that has no option at least to cancel crosstalk between the simulated loudspeakers. It's mainly this crosstalk that puts MIR at a disadvantage, quality-wise.

C'est la vie.

[richhickey]

@Dietz said:

Interesting thread, great suggestions - but as I wrote above: There are no plans to implement a proprietary binaural encoder into MIR 3D in the near future.

And yet it's clearly a feature desired by your customers, myself included. Nothing ruins the sound of MIR3D faster than DearVR. Please reconsider.

Thanks,

Rich

[Dietz]

@richhickey said:

[...] Nothing ruins the sound of MIR3D faster than DearVR.

How comes? Would you mind to explain? Thx!

[richhickey]

@richhickey said:

[...] Nothing ruins the sound of MIR3D faster than DearVR.

How comes? Would you mind to explain? Thx!

I find the sound of MIR 3D through DearVR to be somewhat dulled (in frequency) and blurrier (in time) than without, and it doesn't yield a compelling binaural experience in the end. E.g. I find DearVR can't present sounds as coming from in front of you very well at all (they end up 'in your head' just as do ordinary centered sounds through headphones), compared to a real binaural recording which does that convincingly. It's subjective and YMMV

[Dietz]

I see. But there is not much more I can add to the topic other than quoting my response to your request in VI Control's Forum:

@Another User said:

[...] Let's face it: Choosing a "good" binauralisation algo is as controversial as the quest for the "best legato" in string libraries these days.  It all depends on personal expectations, habits, and most of all: the physical shapes of heads and ears. As the most obvious example let me point out that I actually happen to like DearVR's binauralisation a lot.  ... in the end it's a somewhat pointless discussion anyway, considering that commercial releases will have to rely on Dolby's or Apple's algorithms most of the time, like it or not.

Bottom line: Never say never, but I don't expect VSL developers to enter this minefield anytime soon, especially given the fact that there are many useful options available - some of them even for free.

Thanks for your understanding!